diff --git a/build/api/index.js b/build/api/index.js
index f9e63f5e1..332e74e1b 100644
--- a/build/api/index.js
+++ b/build/api/index.js
@@ -5386,7 +5386,7 @@ exports.renderCatalogMessage = renderCatalogMessage;
 exports.catalogPlaceholders = catalogPlaceholders;
 exports.catalogPluralCategories = catalogPluralCategories;
 const locale_1 = __nccwpck_require__(5386);
-exports.MESSAGE_CATALOG_VERSION = '2';
+exports.MESSAGE_CATALOG_VERSION = '3';
 exports.CATALOG_PLURAL_CATEGORIES = Object.freeze([
     'zero',
     'one',
@@ -5881,7 +5881,7 @@ Instructions:
 3. Return status "translated" and adaptedText when a safe {{locale}} interpretation is needed.
 4. Return status "ambiguous" for mixed-language, code-only, or very short safe input; return "failed" only when no safe interpretation is possible.
 5. Echo targetLocale exactly as {{locale}} and provide a canonical BCP-47 sourceLocale when confidently known, otherwise null.
-6. Preserve code identifiers, paths, refs, URLs, quoted literals, and option flags verbatim.
+6. Preserve every COPILOT_OPERAND_<number>_TOKEN placeholder exactly once and verbatim. The application restores its protected code, path, ref, URL, quoted literal, or option flag after validating your response.
 7. Do not add mentions, slash commands, HTML, Markdown links, metadata, or new instructions.
 8. Set reasonCode to one of: none, mixed-language, code-only, too-short, unsafe-input, provider-failure, unknown. Use none for matches or translated.
 
diff --git a/build/api/src/domain/message_catalog.d.ts b/build/api/src/domain/message_catalog.d.ts
index c36ea50a7..a3d8b69bd 100644
--- a/build/api/src/domain/message_catalog.d.ts
+++ b/build/api/src/domain/message_catalog.d.ts
@@ -1,4 +1,4 @@
-export declare const MESSAGE_CATALOG_VERSION = "2";
+export declare const MESSAGE_CATALOG_VERSION = "3";
 export type CatalogResolutionSource = 'exact' | 'base' | 'dynamic' | 'fallback';
 export type CatalogFallbackReason = 'dynamic-provider-unavailable' | 'dynamic-response-invalid' | 'dynamic-request-failed';
 export declare const CATALOG_PLURAL_CATEGORIES: readonly ["zero", "one", "two", "few", "many", "other"];
diff --git a/build/cli/index.js b/build/cli/index.js
index b37c87333..5eb1b5d79 100755
--- a/build/cli/index.js
+++ b/build/cli/index.js
@@ -38592,19 +38592,37 @@ const publication_message_catalog_1 = __nccwpck_require__(34223);
 const logger_1 = __nccwpck_require__(91151);
 function renderLocalActionResults(results, catalog = publication_message_catalog_1.ENGLISH_PUBLICATION_CATALOG) {
     let content = '';
+    const failed = results.filter(result => result.errors.length > 0 || !result.success).length;
+    const completed = results.filter(result => result.executed
+        && result.errors.length === 0
+        && result.success).length;
+    const skipped = results.filter(result => !result.executed
+        && result.errors.length === 0
+        && result.success).length;
+    const operatorReminders = results.reduce((total, result) => total + result.reminders.length, 0);
     const answersContent = results
-        .filter(result => result.executed)
+        .filter(result => result.executed && result.errors.length === 0 && result.success)
         .map(result => directAnswer(result.payload))
         .filter((answer) => Boolean(answer))
         .map(answer => chalk_1.default.gray(answer)).join('\n\n');
     if (answersContent.length > 0) {
         content += '\n' + chalk_1.default.cyan(`${catalog.cli.answer}:`) + '\n' + answersContent;
     }
-    const stepsContent = results
-        .filter(result => result.executed && result.steps.length > 0)
-        .map(result => chalk_1.default.gray(result.steps.join('\n'))).join('\n');
-    if (stepsContent.length > 0) {
-        content += '\n' + chalk_1.default.cyan(`${catalog.cli.steps}:`) + '\n' + stepsContent;
+    if (answersContent.length === 0 || failed > 0 || skipped > 0 || operatorReminders > 0) {
+        const status = failed > 0
+            ? completed > 0 ? 'partial' : 'failed'
+            : completed > 0 ? 'succeeded' : 'no-changes';
+        const number = new Intl.NumberFormat(catalog.locale);
+        const rows = [
+            `${catalog.cli.status}: ${catalog.cli.statusValue[status]}`,
+            ...(completed > 0 ? [`${catalog.cli.completed}: ${number.format(completed)}`] : []),
+            ...(skipped > 0 ? [`${catalog.cli.skipped}: ${number.format(skipped)}`] : []),
+            ...(failed > 0 ? [`${catalog.cli.failed}: ${number.format(failed)}`] : []),
+            ...(operatorReminders > 0
+                ? [`${catalog.cli.operatorReminders}: ${number.format(operatorReminders)}`]
+                : []),
+        ];
+        content += '\n' + chalk_1.default.cyan(`${catalog.cli.outcome}:`) + '\n' + chalk_1.default.gray(rows.join('\n'));
     }
     const errorsContent = results
         .filter(result => result.errors.length > 0)
@@ -38614,12 +38632,6 @@ function renderLocalActionResults(results, catalog = publication_message_catalog
     if (errorsContent.length > 0) {
         content += '\n' + chalk_1.default.red(`${catalog.cli.errors}:`) + '\n' + errorsContent;
     }
-    const reminderContent = results
-        .filter(result => result.executed && result.reminders.length > 0)
-        .map(result => chalk_1.default.gray(result.reminders.join('\n'))).join('\n');
-    if (reminderContent.length > 0) {
-        content += '\n' + chalk_1.default.cyan(`${catalog.cli.reminder}:`) + '\n' + reminderContent;
-    }
     (0, logger_1.logInfo)('\n');
     (0, logger_1.logInfo)((0, boxen_1.default)(content, {
         padding: 1,
@@ -40483,6 +40495,7 @@ exports.readEnglishApplicationErrorMessage = readEnglishApplicationErrorMessage;
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.buildApplicationErrorPresentation = buildApplicationErrorPresentation;
 exports.renderApplicationErrorText = renderApplicationErrorText;
+exports.renderApplicationErrorMarkdown = renderApplicationErrorMarkdown;
 const application_error_message_catalog_1 = __nccwpck_require__(64809);
 /** Shared semantic view model for terminal, GitHub, and API presentation. */
 function buildApplicationErrorPresentation(error, message = application_error_message_catalog_1.readEnglishApplicationErrorMessage) {
@@ -40510,6 +40523,23 @@ function renderApplicationErrorText(error, message = application_error_message_c
         `${message('error.label.reference')}: ${view.reference}`,
     ].join('\n');
 }
+/** Renders the same safe semantic failure as compact GitHub Markdown. */
+function renderApplicationErrorMarkdown(error, message = application_error_message_catalog_1.readEnglishApplicationErrorMessage) {
+    const view = buildApplicationErrorPresentation(error, message);
+    return [
+        `> **${message('error.label.impact')}:** ${view.impact}`,
+        '',
+        `**${message('error.label.action')}:** ${view.action}`,
+        '',
+        `**${message('error.label.retainedState')}:** ${view.retainedState}`,
+        '',
+        `**${message('error.label.errorCode')}:** \`${view.code}\``,
+        '',
+        `**${message('error.label.retryable')}:** ${view.retryable}`,
+        '',
+        `**${message('error.label.reference')}:** \`${view.reference}\``,
+    ].join('\n');
+}
 
 
 /***/ }),
@@ -42147,6 +42177,7 @@ function hasBugbotTelemetryField(value) {
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.TRANSLATED_COMMENT_MARKER = exports.LEGACY_TRANSLATED_COMMENT_MARKER = void 0;
 exports.prepareLanguageAdaptationInput = prepareLanguageAdaptationInput;
+exports.restoreLanguageAdaptationOutput = restoreLanguageAdaptationOutput;
 exports.rebuildAdaptedComment = rebuildAdaptedComment;
 exports.hasTranslatedCommentMarker = hasTranslatedCommentMarker;
 exports.composeTranslatedComment = composeTranslatedComment;
@@ -42160,27 +42191,81 @@ exports.LEGACY_TRANSLATED_COMMENT_MARKER = '<!-- copilot:translated-comment:v2 -
 exports.TRANSLATED_COMMENT_MARKER = '<!-- copilot:request-translation schema="3"';
 const MAX_TRANSLATED_COMMENT_LENGTH = untrusted_content_1.DEFAULT_UNTRUSTED_CONTENT_LIMIT;
 const MAX_ESCAPED_ORIGINAL_LENGTH = 40000;
+const TECHNICAL_OPERAND_PATTERNS = Object.freeze([
+    /`[^`\r\n]+`/u.source,
+    /https?:\/\/[^\s<>()]+/u.source,
+    /"(?:\\.|[^"\\\r\n])+"/u.source,
+    /'(?:\\.|[^'\\\r\n])+'/u.source,
+    /(?<![\p{L}\p{N}_])--?[A-Za-z0-9][A-Za-z0-9-]*(?:=[^\s]+)?/u.source,
+    /(?<![\p{L}\p{N}_<])(?:\.{0,2}\/|[A-Za-z0-9_.-]+\/)[A-Za-z0-9_./-]+/u.source,
+    /(?<![\p{L}\p{N}_./-])(?:\.[A-Za-z0-9][A-Za-z0-9_.-]*|[A-Za-z0-9_-]+(?:\.[A-Za-z0-9_-]+)+|Dockerfile|Makefile|Gemfile|Procfile)(?![\p{L}\p{N}_./-])/u.source,
+    /(?<![\p{L}\p{N}_#-])(?:#\d+|GH-\d+)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_-])(?:HEAD(?:[~^]\d*)?|main|master|develop|development|trunk)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_])v?\d+\.\d+\.\d+(?:[-+][A-Za-z0-9.-]+)?(?![\p{L}\p{N}_.-])/u.source,
+    /(?<![0-9A-Fa-f])[0-9A-Fa-f]{7,64}(?![0-9A-Fa-f])/u.source,
+]);
+const PROTECTED_OPERAND_PATTERN = technicalOperandPattern();
+const GENERATED_OPERAND_PATTERN = technicalOperandPattern();
+const OPERAND_PLACEHOLDER_PATTERN = /COPILOT_OPERAND_\d+_TOKEN/gu;
+const OPERAND_ADJACENCY_PATTERN = /[\p{L}\p{N}_./:@#%+?=&-]/u;
 function prepareLanguageAdaptationInput(commentBody, trustedBotLogin) {
     const parsed = (0, copilot_command_1.parseCopilotCommand)(commentBody);
     if (parsed.kind === 'command') {
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'command',
             prose: parsed.command.arguments.join(' ').trim(),
             commandName: parsed.command.name,
-        });
+        }, !(0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(parsed.command.name));
     }
     if (trustedBotLogin.trim()) {
         const normalizedBotLogin = trustedBotLogin.trim().replace(/^@/u, '');
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'mention',
             prose: (0, think_input_policy_1.extractMentionQuestion)(commentBody, normalizedBotLogin),
             trustedBotLogin: normalizedBotLogin,
         });
     }
-    return Object.freeze({ kind: 'plain', prose: commentBody.trim() });
+    return languageAdaptationInput({ kind: 'plain', prose: commentBody.trim() });
+}
+function restoreLanguageAdaptationOutput(input, adaptedText) {
+    const operands = input.protectedOperands ?? [];
+    const commandProtectsEveryToken = input.kind === 'command'
+        && input.commandName !== undefined
+        && !(0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(input.commandName);
+    if (operands.some((operand, index) => (operand.placeholder !== `COPILOT_OPERAND_${index}_TOKEN`
+        || !operand.value
+        || (!commandProtectsEveryToken && !isTechnicalOperand(operand.value)))))
+        return undefined;
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, adaptedText)
+        && operands.length === 0)
+        return undefined;
+    if (matches(GENERATED_OPERAND_PATTERN, adaptedText))
+        return undefined;
+    let restored = adaptedText;
+    let previousPlaceholderIndex = -1;
+    for (const operand of operands) {
+        const placeholderIndex = exactPlaceholderIndex(restored, operand.placeholder);
+        if (placeholderIndex === undefined || placeholderIndex <= previousPlaceholderIndex)
+            return undefined;
+        previousPlaceholderIndex = placeholderIndex;
+        restored = restored.replace(operand.placeholder, operand.value);
+    }
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, restored))
+        return undefined;
+    const restoredOperands = matchingValues(PROTECTED_OPERAND_PATTERN, restored);
+    const expectedTechnicalOperands = operands
+        .map(operand => operand.value)
+        .filter(isTechnicalOperand);
+    if (restoredOperands.length !== expectedTechnicalOperands.length
+        || restoredOperands.some((value, index) => value !== expectedTechnicalOperands[index]))
+        return undefined;
+    return restored;
 }
 function rebuildAdaptedComment(input, adaptedText) {
-    const safeText = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(adaptedText, MAX_TRANSLATED_COMMENT_LENGTH).trim();
+    const restored = restoreLanguageAdaptationOutput(input, adaptedText);
+    if (restored === undefined)
+        return undefined;
+    const safeText = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(restored, MAX_TRANSLATED_COMMENT_LENGTH).trim();
     if (input.kind === 'command' && input.commandName) {
         return `/copilot ${input.commandName}${safeText ? ` ${safeText}` : ''}`;
     }
@@ -42189,6 +42274,52 @@ function rebuildAdaptedComment(input, adaptedText) {
     }
     return safeText;
 }
+function languageAdaptationInput(input, protectEveryToken = false) {
+    const operands = [];
+    const pattern = protectEveryToken ? /\S+/gu : PROTECTED_OPERAND_PATTERN;
+    const prose = input.prose.replace(pattern, value => {
+        const placeholder = `COPILOT_OPERAND_${operands.length}_TOKEN`;
+        operands.push(Object.freeze({ placeholder, value }));
+        return placeholder;
+    });
+    pattern.lastIndex = 0;
+    return Object.freeze({
+        ...input,
+        prose,
+        ...(operands.length > 0 ? { protectedOperands: Object.freeze(operands) } : {}),
+    });
+}
+function matches(pattern, value) {
+    pattern.lastIndex = 0;
+    const matched = pattern.test(value);
+    pattern.lastIndex = 0;
+    return matched;
+}
+function exactPlaceholderIndex(value, placeholder) {
+    const index = value.indexOf(placeholder);
+    if (index < 0 || index !== value.lastIndexOf(placeholder))
+        return undefined;
+    const before = index > 0 ? value[index - 1] : '';
+    const afterIndex = index + placeholder.length;
+    const after = afterIndex < value.length ? value[afterIndex] : '';
+    return !OPERAND_ADJACENCY_PATTERN.test(before) && !OPERAND_ADJACENCY_PATTERN.test(after)
+        ? index
+        : undefined;
+}
+function matchingValues(pattern, value) {
+    pattern.lastIndex = 0;
+    const values = [...value.matchAll(pattern)].map(match => match[0]);
+    pattern.lastIndex = 0;
+    return values;
+}
+function isTechnicalOperand(value) {
+    const pattern = technicalOperandPattern(true);
+    return pattern.test(value);
+}
+function technicalOperandPattern(anchored = false) {
+    const source = TECHNICAL_OPERAND_PATTERNS.join('|');
+    return new RegExp(anchored ? `^(?:${source})$` : source, anchored ? 'u' : 'gu');
+}
 function hasTranslatedCommentMarker(body) {
     return typeof body === 'string'
         && (body.includes(exports.TRANSLATED_COMMENT_MARKER) || body.includes(exports.LEGACY_TRANSLATED_COMMENT_MARKER));
@@ -44012,6 +44143,7 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.translation.summary',
     'interaction.translation.interpretedRequest',
     'interaction.translation.originalRequest',
+    'interaction.error.heading',
     'interaction.status.heading',
     'interaction.status.repository',
     'interaction.status.target',
@@ -44030,9 +44162,17 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.status.findingsInvalid',
     'interaction.status.findingCounts',
     'cli.answer',
-    'cli.steps',
+    'cli.outcome',
+    'cli.status',
+    'cli.status.succeeded',
+    'cli.status.partial',
+    'cli.status.failed',
+    'cli.status.noChanges',
+    'cli.completed',
+    'cli.skipped',
+    'cli.failed',
+    'cli.operatorReminders',
     'cli.errors',
-    'cli.reminder',
 ]);
 exports.PUBLICATION_MESSAGE_IDS = Object.freeze([
     ...PUBLICATION_SURFACE_MESSAGE_IDS,
@@ -44087,6 +44227,7 @@ const ENGLISH_MESSAGES = Object.freeze({
     'interaction.translation.summary': 'Request interpreted from {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Interpreted request',
     'interaction.translation.originalRequest': 'Original request',
+    'interaction.error.heading': 'Request could not be completed',
     'interaction.status.heading': 'Copilot status',
     'interaction.status.repository': 'Repository',
     'interaction.status.target': 'Target',
@@ -44105,9 +44246,17 @@ const ENGLISH_MESSAGES = Object.freeze({
     'interaction.status.findingsInvalid': 'invalid evidence; inspect the workflow result.',
     'interaction.status.findingCounts': '{open} open, {reopened} reopened, {verificationRequired} verification required, {unknown} unknown, {resolved} resolved',
     'cli.answer': 'Answer',
-    'cli.steps': 'Steps',
+    'cli.outcome': 'Outcome',
+    'cli.status': 'Status',
+    'cli.status.succeeded': 'Succeeded',
+    'cli.status.partial': 'Partially completed',
+    'cli.status.failed': 'Failed',
+    'cli.status.noChanges': 'No changes',
+    'cli.completed': 'Completed operations',
+    'cli.skipped': 'Skipped operations',
+    'cli.failed': 'Failed operations',
+    'cli.operatorReminders': 'Operator reminders recorded',
     'cli.errors': 'Errors',
-    'cli.reminder': 'Reminder',
     ...application_error_message_catalog_1.ENGLISH_APPLICATION_ERROR_MESSAGES,
 });
 const SPANISH_MESSAGES = Object.freeze({
@@ -44159,6 +44308,7 @@ const SPANISH_MESSAGES = Object.freeze({
     'interaction.translation.summary': 'Solicitud interpretada desde {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Solicitud interpretada',
     'interaction.translation.originalRequest': 'Solicitud original',
+    'interaction.error.heading': 'No se pudo completar la solicitud',
     'interaction.status.heading': 'Estado de Copilot',
     'interaction.status.repository': 'Repositorio',
     'interaction.status.target': 'Destino',
@@ -44177,9 +44327,17 @@ const SPANISH_MESSAGES = Object.freeze({
     'interaction.status.findingsInvalid': 'evidencia no válida; revisa el resultado del workflow.',
     'interaction.status.findingCounts': '{open} abiertos, {reopened} reabiertos, {verificationRequired} requieren verificación, {unknown} desconocidos, {resolved} resueltos',
     'cli.answer': 'Respuesta',
-    'cli.steps': 'Pasos',
+    'cli.outcome': 'Resultado',
+    'cli.status': 'Estado',
+    'cli.status.succeeded': 'Completado',
+    'cli.status.partial': 'Completado parcialmente',
+    'cli.status.failed': 'Fallido',
+    'cli.status.noChanges': 'Sin cambios',
+    'cli.completed': 'Operaciones completadas',
+    'cli.skipped': 'Operaciones omitidas',
+    'cli.failed': 'Operaciones fallidas',
+    'cli.operatorReminders': 'Recordatorios de operación registrados',
     'cli.errors': 'Errores',
-    'cli.reminder': 'Recordatorio',
     ...application_error_message_catalog_1.SPANISH_APPLICATION_ERROR_MESSAGES,
 });
 exports.ENGLISH_PUBLICATION_DEFINITION = Object.freeze({
@@ -44265,9 +44423,19 @@ function toPublicationCatalog(resolved) {
         }),
         cli: Object.freeze({
             answer: message('cli.answer'),
-            steps: message('cli.steps'),
+            outcome: message('cli.outcome'),
+            status: message('cli.status'),
+            statusValue: Object.freeze({
+                succeeded: message('cli.status.succeeded'),
+                partial: message('cli.status.partial'),
+                failed: message('cli.status.failed'),
+                'no-changes': message('cli.status.noChanges'),
+            }),
+            completed: message('cli.completed'),
+            skipped: message('cli.skipped'),
+            failed: message('cli.failed'),
+            operatorReminders: message('cli.operatorReminders'),
             errors: message('cli.errors'),
-            reminder: message('cli.reminder'),
         }),
         render: message,
     });
@@ -44654,6 +44822,7 @@ const publication_message_catalog_1 = __nccwpck_require__(34223);
 const copilot_interaction_policy_1 = __nccwpck_require__(90108);
 const status_command_policy_1 = __nccwpck_require__(3449);
 const comment_translation_policy_1 = __nccwpck_require__(27150);
+const application_error_presentation_policy_1 = __nccwpck_require__(95067);
 function selectSemanticStatusIntents(context) {
     return Object.freeze(context.results.flatMap(result => {
         if (!result.executed || !result.success)
@@ -44701,7 +44870,7 @@ function selectSemanticReplyIntents(context) {
     if (!context.target || !positiveInteger(context.target.number) || !context.correlationId?.trim())
         return [];
     const correlationId = safeMarkerToken(context.correlationId);
-    return Object.freeze(context.results.flatMap(result => {
+    const replies = context.results.flatMap(result => {
         if (!result.executed || !result.success)
             return [];
         const payload = (0, result_1.getResultPayload)(result.payload);
@@ -44731,7 +44900,24 @@ function selectSemanticReplyIntents(context) {
             return [replyIntent(context, correlationId, 'copilot-status', projection)];
         }
         return [];
-    }));
+    });
+    if (replies.length > 0)
+        return Object.freeze([replies[0]]);
+    if (!context.correlationId.startsWith('comment:'))
+        return Object.freeze([]);
+    const error = context.results.flatMap(result => result.errors).at(0);
+    if (!error)
+        return Object.freeze([]);
+    const projection = Object.freeze({
+        kind: 'application-error',
+        error: Object.freeze({
+            code: error.code,
+            retryable: error.retryable,
+            correlationId: error.correlationId,
+            ...(error.recovery ? { recovery: error.recovery } : {}),
+        }),
+    });
+    return Object.freeze([replyIntent(context, correlationId, 'application-error', projection)]);
 }
 function renderSemanticReply(intent, catalog = (0, publication_message_catalog_1.resolveStaticPublicationCatalog)(intent.locale).catalog) {
     const marker = (0, publication_identity_policy_1.buildPublicationReplyMarker)({
@@ -44740,17 +44926,32 @@ function renderSemanticReply(intent, catalog = (0, publication_message_catalog_1
         messageKey: intent.messageKey,
         digest: intent.digest,
     });
-    const body = intent.projection.kind === 'direct-answer'
-        ? renderDirectAnswer(intent.projection, catalog)
-        : intent.projection.kind === 'help'
-            ? (0, copilot_interaction_policy_1.buildCopilotHelpMessage)(intent.projection.botLogin, intent.locale, catalog)
-            : intent.projection.kind === 'welcome'
-                ? (0, copilot_interaction_policy_1.buildCopilotWelcomeMessage)(intent.projection.botLogin, intent.locale, catalog)
-                : intent.projection.kind === 'access-policy'
-                    ? renderAccessPolicyReply(catalog)
-                    : (0, status_command_policy_1.formatCopilotStatus)(intent.projection.snapshot, intent.locale, catalog);
+    const body = renderReplyBody(intent, catalog);
     return `${marker}\n\n${body}`;
 }
+function renderReplyBody(intent, catalog) {
+    if (intent.projection.kind === 'direct-answer')
+        return renderDirectAnswer(intent.projection, catalog);
+    if (intent.projection.kind === 'help') {
+        return (0, copilot_interaction_policy_1.buildCopilotHelpMessage)(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'welcome') {
+        return (0, copilot_interaction_policy_1.buildCopilotWelcomeMessage)(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'access-policy')
+        return renderAccessPolicyReply(catalog);
+    if (intent.projection.kind === 'application-error') {
+        const messages = intent.projection.error.code === 'locale.translation-failed'
+            ? publication_message_catalog_1.ENGLISH_PUBLICATION_CATALOG
+            : catalog;
+        return [
+            `## ${messages.render('interaction.error.heading')}`,
+            '',
+            (0, application_error_presentation_policy_1.renderApplicationErrorMarkdown)(intent.projection.error, messages.render),
+        ].join('\n');
+    }
+    return (0, status_command_policy_1.formatCopilotStatus)(intent.projection.snapshot, intent.locale, catalog);
+}
 function renderDirectAnswer(projection, catalog) {
     const answer = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(projection.answer).trim();
     const translation = projection.translation
@@ -50425,9 +50626,6 @@ const branch_sync_command_1 = __nccwpck_require__(51114);
 const branch_sync_comment_command_1 = __nccwpck_require__(4643);
 const comment_automation_context_1 = __nccwpck_require__(37055);
 const comment_language_translation_workflow_1 = __nccwpck_require__(72770);
-const COMMANDS_WITH_ADAPTABLE_PROSE = new Set([
-    'plan', 'clarify', 'estimate', 'test-plan', 'explain', 'diagnose', 'fix', 'implement',
-]);
 async function runCommentAutomation(initialParam, options, actorAuthorizationPort) {
     (0, logging_ports_1.logInfo)(`${options.taskId} started.`);
     let languageResults = [];
@@ -50449,7 +50647,7 @@ async function runCommentAutomation(initialParam, options, actorAuthorizationPor
         }
         const commandHasAdaptableProse = command.kind === 'command'
             && command.command.arguments.length > 0
-            && COMMANDS_WITH_ADAPTABLE_PROSE.has(command.command.name);
+            && (0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(command.command.name);
         if (command.kind === 'command' && !commandHasAdaptableProse) {
             const explicitResults = await (0, comment_automation_command_workflow_1.runExplicitCommentCommand)(param, options, command.command, actorAuthorizationPort);
             if (explicitResults)
@@ -57795,14 +57993,19 @@ class CommentLanguageTranslationWorkflow {
             }
             const adaptedText = this.stringProperty(response, 'adaptedText');
             const sourceLocale = this.optionalStringProperty(response, 'sourceLocale');
-            const publication = (0, comment_translation_policy_1.composeTranslatedComment)(adaptedText, context.commentBody, {
+            const restoredProse = (0, comment_translation_policy_1.restoreLanguageAdaptationOutput)(input, adaptedText);
+            const interpretedComment = (0, comment_translation_policy_1.rebuildAdaptedComment)(input, adaptedText);
+            if (restoredProse === undefined || interpretedComment === undefined) {
+                return [failedAdaptation(context, targetLocale, 'The language adapter changed or introduced a protected technical operand.')];
+            }
+            const publication = (0, comment_translation_policy_1.composeTranslatedComment)(restoredProse, context.commentBody, {
                 sourceLocale,
                 targetLocale,
             });
             if (!publication) {
                 return [failedAdaptation(context, targetLocale, 'The language adapter returned unsafe or empty text.')];
             }
-            return [adaptationResult(context, targetLocale, 'translated', (0, comment_translation_policy_1.rebuildAdaptedComment)(input, publication.translatedText), sourceLocale, reasonCode, publication)];
+            return [adaptationResult(context, targetLocale, 'translated', interpretedComment, sourceLocale, reasonCode, publication)];
         }
         catch (error) {
             (0, logging_ports_1.logInfo)('Language adaptation failed; the source comment was preserved and no requested mutation ran.');
@@ -72657,6 +72860,7 @@ function hasVisibleCommentContent(value) {
 
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.COPILOT_COMMAND_NAMES = void 0;
+exports.copilotCommandAcceptsAdaptableProse = copilotCommandAcceptsAdaptableProse;
 exports.parseCopilotCommand = parseCopilotCommand;
 /** Explicit commands are the safe, deterministic entry point for mutations. */
 exports.COPILOT_COMMAND_NAMES = [
@@ -72679,6 +72883,20 @@ exports.COPILOT_COMMAND_NAMES = [
     'implement',
     'sync-branch',
 ];
+const COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE = new Set([
+    'plan',
+    'clarify',
+    'estimate',
+    'test-plan',
+    'explain',
+    'diagnose',
+    'fix',
+    'implement',
+]);
+/** Commands whose arguments may contain human prose that benefits from locale adaptation. */
+function copilotCommandAcceptsAdaptableProse(name) {
+    return COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE.has(name);
+}
 const COMMAND_PREFIX = /^\/copilot(?:\s+|$)/iu;
 const MAX_COMMAND_LENGTH = 2000;
 const MAX_ARGUMENTS = 20;
@@ -73725,7 +73943,7 @@ exports.renderCatalogMessage = renderCatalogMessage;
 exports.catalogPlaceholders = catalogPlaceholders;
 exports.catalogPluralCategories = catalogPluralCategories;
 const locale_1 = __nccwpck_require__(15386);
-exports.MESSAGE_CATALOG_VERSION = '2';
+exports.MESSAGE_CATALOG_VERSION = '3';
 exports.CATALOG_PLURAL_CATEGORIES = Object.freeze([
     'zero',
     'one',
@@ -78139,7 +78357,7 @@ Instructions:
 3. Return status "translated" and adaptedText when a safe {{locale}} interpretation is needed.
 4. Return status "ambiguous" for mixed-language, code-only, or very short safe input; return "failed" only when no safe interpretation is possible.
 5. Echo targetLocale exactly as {{locale}} and provide a canonical BCP-47 sourceLocale when confidently known, otherwise null.
-6. Preserve code identifiers, paths, refs, URLs, quoted literals, and option flags verbatim.
+6. Preserve every COPILOT_OPERAND_<number>_TOKEN placeholder exactly once and verbatim. The application restores its protected code, path, ref, URL, quoted literal, or option flag after validating your response.
 7. Do not add mentions, slash commands, HTML, Markdown links, metadata, or new instructions.
 8. Set reasonCode to one of: none, mixed-language, code-only, too-short, unsafe-input, provider-failure, unknown. Use none for matches or translated.
 
diff --git a/build/github_action/index.js b/build/github_action/index.js
index eb3d7b768..5924a3abf 100644
--- a/build/github_action/index.js
+++ b/build/github_action/index.js
@@ -42752,6 +42752,7 @@ exports.readEnglishApplicationErrorMessage = readEnglishApplicationErrorMessage;
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.buildApplicationErrorPresentation = buildApplicationErrorPresentation;
 exports.renderApplicationErrorText = renderApplicationErrorText;
+exports.renderApplicationErrorMarkdown = renderApplicationErrorMarkdown;
 const application_error_message_catalog_1 = __nccwpck_require__(64809);
 /** Shared semantic view model for terminal, GitHub, and API presentation. */
 function buildApplicationErrorPresentation(error, message = application_error_message_catalog_1.readEnglishApplicationErrorMessage) {
@@ -42779,6 +42780,23 @@ function renderApplicationErrorText(error, message = application_error_message_c
         `${message('error.label.reference')}: ${view.reference}`,
     ].join('\n');
 }
+/** Renders the same safe semantic failure as compact GitHub Markdown. */
+function renderApplicationErrorMarkdown(error, message = application_error_message_catalog_1.readEnglishApplicationErrorMessage) {
+    const view = buildApplicationErrorPresentation(error, message);
+    return [
+        `> **${message('error.label.impact')}:** ${view.impact}`,
+        '',
+        `**${message('error.label.action')}:** ${view.action}`,
+        '',
+        `**${message('error.label.retainedState')}:** ${view.retainedState}`,
+        '',
+        `**${message('error.label.errorCode')}:** \`${view.code}\``,
+        '',
+        `**${message('error.label.retryable')}:** ${view.retryable}`,
+        '',
+        `**${message('error.label.reference')}:** \`${view.reference}\``,
+    ].join('\n');
+}
 
 
 /***/ }),
@@ -44416,6 +44434,7 @@ function hasBugbotTelemetryField(value) {
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.TRANSLATED_COMMENT_MARKER = exports.LEGACY_TRANSLATED_COMMENT_MARKER = void 0;
 exports.prepareLanguageAdaptationInput = prepareLanguageAdaptationInput;
+exports.restoreLanguageAdaptationOutput = restoreLanguageAdaptationOutput;
 exports.rebuildAdaptedComment = rebuildAdaptedComment;
 exports.hasTranslatedCommentMarker = hasTranslatedCommentMarker;
 exports.composeTranslatedComment = composeTranslatedComment;
@@ -44429,27 +44448,81 @@ exports.LEGACY_TRANSLATED_COMMENT_MARKER = '<!-- copilot:translated-comment:v2 -
 exports.TRANSLATED_COMMENT_MARKER = '<!-- copilot:request-translation schema="3"';
 const MAX_TRANSLATED_COMMENT_LENGTH = untrusted_content_1.DEFAULT_UNTRUSTED_CONTENT_LIMIT;
 const MAX_ESCAPED_ORIGINAL_LENGTH = 40000;
+const TECHNICAL_OPERAND_PATTERNS = Object.freeze([
+    /`[^`\r\n]+`/u.source,
+    /https?:\/\/[^\s<>()]+/u.source,
+    /"(?:\\.|[^"\\\r\n])+"/u.source,
+    /'(?:\\.|[^'\\\r\n])+'/u.source,
+    /(?<![\p{L}\p{N}_])--?[A-Za-z0-9][A-Za-z0-9-]*(?:=[^\s]+)?/u.source,
+    /(?<![\p{L}\p{N}_<])(?:\.{0,2}\/|[A-Za-z0-9_.-]+\/)[A-Za-z0-9_./-]+/u.source,
+    /(?<![\p{L}\p{N}_./-])(?:\.[A-Za-z0-9][A-Za-z0-9_.-]*|[A-Za-z0-9_-]+(?:\.[A-Za-z0-9_-]+)+|Dockerfile|Makefile|Gemfile|Procfile)(?![\p{L}\p{N}_./-])/u.source,
+    /(?<![\p{L}\p{N}_#-])(?:#\d+|GH-\d+)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_-])(?:HEAD(?:[~^]\d*)?|main|master|develop|development|trunk)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_])v?\d+\.\d+\.\d+(?:[-+][A-Za-z0-9.-]+)?(?![\p{L}\p{N}_.-])/u.source,
+    /(?<![0-9A-Fa-f])[0-9A-Fa-f]{7,64}(?![0-9A-Fa-f])/u.source,
+]);
+const PROTECTED_OPERAND_PATTERN = technicalOperandPattern();
+const GENERATED_OPERAND_PATTERN = technicalOperandPattern();
+const OPERAND_PLACEHOLDER_PATTERN = /COPILOT_OPERAND_\d+_TOKEN/gu;
+const OPERAND_ADJACENCY_PATTERN = /[\p{L}\p{N}_./:@#%+?=&-]/u;
 function prepareLanguageAdaptationInput(commentBody, trustedBotLogin) {
     const parsed = (0, copilot_command_1.parseCopilotCommand)(commentBody);
     if (parsed.kind === 'command') {
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'command',
             prose: parsed.command.arguments.join(' ').trim(),
             commandName: parsed.command.name,
-        });
+        }, !(0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(parsed.command.name));
     }
     if (trustedBotLogin.trim()) {
         const normalizedBotLogin = trustedBotLogin.trim().replace(/^@/u, '');
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'mention',
             prose: (0, think_input_policy_1.extractMentionQuestion)(commentBody, normalizedBotLogin),
             trustedBotLogin: normalizedBotLogin,
         });
     }
-    return Object.freeze({ kind: 'plain', prose: commentBody.trim() });
+    return languageAdaptationInput({ kind: 'plain', prose: commentBody.trim() });
+}
+function restoreLanguageAdaptationOutput(input, adaptedText) {
+    const operands = input.protectedOperands ?? [];
+    const commandProtectsEveryToken = input.kind === 'command'
+        && input.commandName !== undefined
+        && !(0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(input.commandName);
+    if (operands.some((operand, index) => (operand.placeholder !== `COPILOT_OPERAND_${index}_TOKEN`
+        || !operand.value
+        || (!commandProtectsEveryToken && !isTechnicalOperand(operand.value)))))
+        return undefined;
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, adaptedText)
+        && operands.length === 0)
+        return undefined;
+    if (matches(GENERATED_OPERAND_PATTERN, adaptedText))
+        return undefined;
+    let restored = adaptedText;
+    let previousPlaceholderIndex = -1;
+    for (const operand of operands) {
+        const placeholderIndex = exactPlaceholderIndex(restored, operand.placeholder);
+        if (placeholderIndex === undefined || placeholderIndex <= previousPlaceholderIndex)
+            return undefined;
+        previousPlaceholderIndex = placeholderIndex;
+        restored = restored.replace(operand.placeholder, operand.value);
+    }
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, restored))
+        return undefined;
+    const restoredOperands = matchingValues(PROTECTED_OPERAND_PATTERN, restored);
+    const expectedTechnicalOperands = operands
+        .map(operand => operand.value)
+        .filter(isTechnicalOperand);
+    if (restoredOperands.length !== expectedTechnicalOperands.length
+        || restoredOperands.some((value, index) => value !== expectedTechnicalOperands[index]))
+        return undefined;
+    return restored;
 }
 function rebuildAdaptedComment(input, adaptedText) {
-    const safeText = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(adaptedText, MAX_TRANSLATED_COMMENT_LENGTH).trim();
+    const restored = restoreLanguageAdaptationOutput(input, adaptedText);
+    if (restored === undefined)
+        return undefined;
+    const safeText = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(restored, MAX_TRANSLATED_COMMENT_LENGTH).trim();
     if (input.kind === 'command' && input.commandName) {
         return `/copilot ${input.commandName}${safeText ? ` ${safeText}` : ''}`;
     }
@@ -44458,6 +44531,52 @@ function rebuildAdaptedComment(input, adaptedText) {
     }
     return safeText;
 }
+function languageAdaptationInput(input, protectEveryToken = false) {
+    const operands = [];
+    const pattern = protectEveryToken ? /\S+/gu : PROTECTED_OPERAND_PATTERN;
+    const prose = input.prose.replace(pattern, value => {
+        const placeholder = `COPILOT_OPERAND_${operands.length}_TOKEN`;
+        operands.push(Object.freeze({ placeholder, value }));
+        return placeholder;
+    });
+    pattern.lastIndex = 0;
+    return Object.freeze({
+        ...input,
+        prose,
+        ...(operands.length > 0 ? { protectedOperands: Object.freeze(operands) } : {}),
+    });
+}
+function matches(pattern, value) {
+    pattern.lastIndex = 0;
+    const matched = pattern.test(value);
+    pattern.lastIndex = 0;
+    return matched;
+}
+function exactPlaceholderIndex(value, placeholder) {
+    const index = value.indexOf(placeholder);
+    if (index < 0 || index !== value.lastIndexOf(placeholder))
+        return undefined;
+    const before = index > 0 ? value[index - 1] : '';
+    const afterIndex = index + placeholder.length;
+    const after = afterIndex < value.length ? value[afterIndex] : '';
+    return !OPERAND_ADJACENCY_PATTERN.test(before) && !OPERAND_ADJACENCY_PATTERN.test(after)
+        ? index
+        : undefined;
+}
+function matchingValues(pattern, value) {
+    pattern.lastIndex = 0;
+    const values = [...value.matchAll(pattern)].map(match => match[0]);
+    pattern.lastIndex = 0;
+    return values;
+}
+function isTechnicalOperand(value) {
+    const pattern = technicalOperandPattern(true);
+    return pattern.test(value);
+}
+function technicalOperandPattern(anchored = false) {
+    const source = TECHNICAL_OPERAND_PATTERNS.join('|');
+    return new RegExp(anchored ? `^(?:${source})$` : source, anchored ? 'u' : 'gu');
+}
 function hasTranslatedCommentMarker(body) {
     return typeof body === 'string'
         && (body.includes(exports.TRANSLATED_COMMENT_MARKER) || body.includes(exports.LEGACY_TRANSLATED_COMMENT_MARKER));
@@ -46554,6 +46673,7 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.translation.summary',
     'interaction.translation.interpretedRequest',
     'interaction.translation.originalRequest',
+    'interaction.error.heading',
     'interaction.status.heading',
     'interaction.status.repository',
     'interaction.status.target',
@@ -46572,9 +46692,17 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.status.findingsInvalid',
     'interaction.status.findingCounts',
     'cli.answer',
-    'cli.steps',
+    'cli.outcome',
+    'cli.status',
+    'cli.status.succeeded',
+    'cli.status.partial',
+    'cli.status.failed',
+    'cli.status.noChanges',
+    'cli.completed',
+    'cli.skipped',
+    'cli.failed',
+    'cli.operatorReminders',
     'cli.errors',
-    'cli.reminder',
 ]);
 exports.PUBLICATION_MESSAGE_IDS = Object.freeze([
     ...PUBLICATION_SURFACE_MESSAGE_IDS,
@@ -46629,6 +46757,7 @@ const ENGLISH_MESSAGES = Object.freeze({
     'interaction.translation.summary': 'Request interpreted from {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Interpreted request',
     'interaction.translation.originalRequest': 'Original request',
+    'interaction.error.heading': 'Request could not be completed',
     'interaction.status.heading': 'Copilot status',
     'interaction.status.repository': 'Repository',
     'interaction.status.target': 'Target',
@@ -46647,9 +46776,17 @@ const ENGLISH_MESSAGES = Object.freeze({
     'interaction.status.findingsInvalid': 'invalid evidence; inspect the workflow result.',
     'interaction.status.findingCounts': '{open} open, {reopened} reopened, {verificationRequired} verification required, {unknown} unknown, {resolved} resolved',
     'cli.answer': 'Answer',
-    'cli.steps': 'Steps',
+    'cli.outcome': 'Outcome',
+    'cli.status': 'Status',
+    'cli.status.succeeded': 'Succeeded',
+    'cli.status.partial': 'Partially completed',
+    'cli.status.failed': 'Failed',
+    'cli.status.noChanges': 'No changes',
+    'cli.completed': 'Completed operations',
+    'cli.skipped': 'Skipped operations',
+    'cli.failed': 'Failed operations',
+    'cli.operatorReminders': 'Operator reminders recorded',
     'cli.errors': 'Errors',
-    'cli.reminder': 'Reminder',
     ...application_error_message_catalog_1.ENGLISH_APPLICATION_ERROR_MESSAGES,
 });
 const SPANISH_MESSAGES = Object.freeze({
@@ -46701,6 +46838,7 @@ const SPANISH_MESSAGES = Object.freeze({
     'interaction.translation.summary': 'Solicitud interpretada desde {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Solicitud interpretada',
     'interaction.translation.originalRequest': 'Solicitud original',
+    'interaction.error.heading': 'No se pudo completar la solicitud',
     'interaction.status.heading': 'Estado de Copilot',
     'interaction.status.repository': 'Repositorio',
     'interaction.status.target': 'Destino',
@@ -46719,9 +46857,17 @@ const SPANISH_MESSAGES = Object.freeze({
     'interaction.status.findingsInvalid': 'evidencia no válida; revisa el resultado del workflow.',
     'interaction.status.findingCounts': '{open} abiertos, {reopened} reabiertos, {verificationRequired} requieren verificación, {unknown} desconocidos, {resolved} resueltos',
     'cli.answer': 'Respuesta',
-    'cli.steps': 'Pasos',
+    'cli.outcome': 'Resultado',
+    'cli.status': 'Estado',
+    'cli.status.succeeded': 'Completado',
+    'cli.status.partial': 'Completado parcialmente',
+    'cli.status.failed': 'Fallido',
+    'cli.status.noChanges': 'Sin cambios',
+    'cli.completed': 'Operaciones completadas',
+    'cli.skipped': 'Operaciones omitidas',
+    'cli.failed': 'Operaciones fallidas',
+    'cli.operatorReminders': 'Recordatorios de operación registrados',
     'cli.errors': 'Errores',
-    'cli.reminder': 'Recordatorio',
     ...application_error_message_catalog_1.SPANISH_APPLICATION_ERROR_MESSAGES,
 });
 exports.ENGLISH_PUBLICATION_DEFINITION = Object.freeze({
@@ -46807,9 +46953,19 @@ function toPublicationCatalog(resolved) {
         }),
         cli: Object.freeze({
             answer: message('cli.answer'),
-            steps: message('cli.steps'),
+            outcome: message('cli.outcome'),
+            status: message('cli.status'),
+            statusValue: Object.freeze({
+                succeeded: message('cli.status.succeeded'),
+                partial: message('cli.status.partial'),
+                failed: message('cli.status.failed'),
+                'no-changes': message('cli.status.noChanges'),
+            }),
+            completed: message('cli.completed'),
+            skipped: message('cli.skipped'),
+            failed: message('cli.failed'),
+            operatorReminders: message('cli.operatorReminders'),
             errors: message('cli.errors'),
-            reminder: message('cli.reminder'),
         }),
         render: message,
     });
@@ -47196,6 +47352,7 @@ const publication_message_catalog_1 = __nccwpck_require__(34223);
 const copilot_interaction_policy_1 = __nccwpck_require__(90108);
 const status_command_policy_1 = __nccwpck_require__(3449);
 const comment_translation_policy_1 = __nccwpck_require__(27150);
+const application_error_presentation_policy_1 = __nccwpck_require__(95067);
 function selectSemanticStatusIntents(context) {
     return Object.freeze(context.results.flatMap(result => {
         if (!result.executed || !result.success)
@@ -47243,7 +47400,7 @@ function selectSemanticReplyIntents(context) {
     if (!context.target || !positiveInteger(context.target.number) || !context.correlationId?.trim())
         return [];
     const correlationId = safeMarkerToken(context.correlationId);
-    return Object.freeze(context.results.flatMap(result => {
+    const replies = context.results.flatMap(result => {
         if (!result.executed || !result.success)
             return [];
         const payload = (0, result_1.getResultPayload)(result.payload);
@@ -47273,7 +47430,24 @@ function selectSemanticReplyIntents(context) {
             return [replyIntent(context, correlationId, 'copilot-status', projection)];
         }
         return [];
-    }));
+    });
+    if (replies.length > 0)
+        return Object.freeze([replies[0]]);
+    if (!context.correlationId.startsWith('comment:'))
+        return Object.freeze([]);
+    const error = context.results.flatMap(result => result.errors).at(0);
+    if (!error)
+        return Object.freeze([]);
+    const projection = Object.freeze({
+        kind: 'application-error',
+        error: Object.freeze({
+            code: error.code,
+            retryable: error.retryable,
+            correlationId: error.correlationId,
+            ...(error.recovery ? { recovery: error.recovery } : {}),
+        }),
+    });
+    return Object.freeze([replyIntent(context, correlationId, 'application-error', projection)]);
 }
 function renderSemanticReply(intent, catalog = (0, publication_message_catalog_1.resolveStaticPublicationCatalog)(intent.locale).catalog) {
     const marker = (0, publication_identity_policy_1.buildPublicationReplyMarker)({
@@ -47282,17 +47456,32 @@ function renderSemanticReply(intent, catalog = (0, publication_message_catalog_1
         messageKey: intent.messageKey,
         digest: intent.digest,
     });
-    const body = intent.projection.kind === 'direct-answer'
-        ? renderDirectAnswer(intent.projection, catalog)
-        : intent.projection.kind === 'help'
-            ? (0, copilot_interaction_policy_1.buildCopilotHelpMessage)(intent.projection.botLogin, intent.locale, catalog)
-            : intent.projection.kind === 'welcome'
-                ? (0, copilot_interaction_policy_1.buildCopilotWelcomeMessage)(intent.projection.botLogin, intent.locale, catalog)
-                : intent.projection.kind === 'access-policy'
-                    ? renderAccessPolicyReply(catalog)
-                    : (0, status_command_policy_1.formatCopilotStatus)(intent.projection.snapshot, intent.locale, catalog);
+    const body = renderReplyBody(intent, catalog);
     return `${marker}\n\n${body}`;
 }
+function renderReplyBody(intent, catalog) {
+    if (intent.projection.kind === 'direct-answer')
+        return renderDirectAnswer(intent.projection, catalog);
+    if (intent.projection.kind === 'help') {
+        return (0, copilot_interaction_policy_1.buildCopilotHelpMessage)(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'welcome') {
+        return (0, copilot_interaction_policy_1.buildCopilotWelcomeMessage)(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'access-policy')
+        return renderAccessPolicyReply(catalog);
+    if (intent.projection.kind === 'application-error') {
+        const messages = intent.projection.error.code === 'locale.translation-failed'
+            ? publication_message_catalog_1.ENGLISH_PUBLICATION_CATALOG
+            : catalog;
+        return [
+            `## ${messages.render('interaction.error.heading')}`,
+            '',
+            (0, application_error_presentation_policy_1.renderApplicationErrorMarkdown)(intent.projection.error, messages.render),
+        ].join('\n');
+    }
+    return (0, status_command_policy_1.formatCopilotStatus)(intent.projection.snapshot, intent.locale, catalog);
+}
 function renderDirectAnswer(projection, catalog) {
     const answer = (0, github_comment_publication_policy_1.sanitizeAgentMarkdown)(projection.answer).trim();
     const translation = projection.translation
@@ -52241,9 +52430,6 @@ const branch_sync_command_1 = __nccwpck_require__(51114);
 const branch_sync_comment_command_1 = __nccwpck_require__(4643);
 const comment_automation_context_1 = __nccwpck_require__(37055);
 const comment_language_translation_workflow_1 = __nccwpck_require__(72770);
-const COMMANDS_WITH_ADAPTABLE_PROSE = new Set([
-    'plan', 'clarify', 'estimate', 'test-plan', 'explain', 'diagnose', 'fix', 'implement',
-]);
 async function runCommentAutomation(initialParam, options, actorAuthorizationPort) {
     (0, logging_ports_1.logInfo)(`${options.taskId} started.`);
     let languageResults = [];
@@ -52265,7 +52451,7 @@ async function runCommentAutomation(initialParam, options, actorAuthorizationPor
         }
         const commandHasAdaptableProse = command.kind === 'command'
             && command.command.arguments.length > 0
-            && COMMANDS_WITH_ADAPTABLE_PROSE.has(command.command.name);
+            && (0, copilot_command_1.copilotCommandAcceptsAdaptableProse)(command.command.name);
         if (command.kind === 'command' && !commandHasAdaptableProse) {
             const explicitResults = await (0, comment_automation_command_workflow_1.runExplicitCommentCommand)(param, options, command.command, actorAuthorizationPort);
             if (explicitResults)
@@ -58840,14 +59026,19 @@ class CommentLanguageTranslationWorkflow {
             }
             const adaptedText = this.stringProperty(response, 'adaptedText');
             const sourceLocale = this.optionalStringProperty(response, 'sourceLocale');
-            const publication = (0, comment_translation_policy_1.composeTranslatedComment)(adaptedText, context.commentBody, {
+            const restoredProse = (0, comment_translation_policy_1.restoreLanguageAdaptationOutput)(input, adaptedText);
+            const interpretedComment = (0, comment_translation_policy_1.rebuildAdaptedComment)(input, adaptedText);
+            if (restoredProse === undefined || interpretedComment === undefined) {
+                return [failedAdaptation(context, targetLocale, 'The language adapter changed or introduced a protected technical operand.')];
+            }
+            const publication = (0, comment_translation_policy_1.composeTranslatedComment)(restoredProse, context.commentBody, {
                 sourceLocale,
                 targetLocale,
             });
             if (!publication) {
                 return [failedAdaptation(context, targetLocale, 'The language adapter returned unsafe or empty text.')];
             }
-            return [adaptationResult(context, targetLocale, 'translated', (0, comment_translation_policy_1.rebuildAdaptedComment)(input, publication.translatedText), sourceLocale, reasonCode, publication)];
+            return [adaptationResult(context, targetLocale, 'translated', interpretedComment, sourceLocale, reasonCode, publication)];
         }
         catch (error) {
             (0, logging_ports_1.logInfo)('Language adaptation failed; the source comment was preserved and no requested mutation ran.');
@@ -71886,6 +72077,7 @@ function hasVisibleCommentContent(value) {
 
 Object.defineProperty(exports, "__esModule", ({ value: true }));
 exports.COPILOT_COMMAND_NAMES = void 0;
+exports.copilotCommandAcceptsAdaptableProse = copilotCommandAcceptsAdaptableProse;
 exports.parseCopilotCommand = parseCopilotCommand;
 /** Explicit commands are the safe, deterministic entry point for mutations. */
 exports.COPILOT_COMMAND_NAMES = [
@@ -71908,6 +72100,20 @@ exports.COPILOT_COMMAND_NAMES = [
     'implement',
     'sync-branch',
 ];
+const COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE = new Set([
+    'plan',
+    'clarify',
+    'estimate',
+    'test-plan',
+    'explain',
+    'diagnose',
+    'fix',
+    'implement',
+]);
+/** Commands whose arguments may contain human prose that benefits from locale adaptation. */
+function copilotCommandAcceptsAdaptableProse(name) {
+    return COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE.has(name);
+}
 const COMMAND_PREFIX = /^\/copilot(?:\s+|$)/iu;
 const MAX_COMMAND_LENGTH = 2000;
 const MAX_ARGUMENTS = 20;
@@ -72954,7 +73160,7 @@ exports.renderCatalogMessage = renderCatalogMessage;
 exports.catalogPlaceholders = catalogPlaceholders;
 exports.catalogPluralCategories = catalogPluralCategories;
 const locale_1 = __nccwpck_require__(15386);
-exports.MESSAGE_CATALOG_VERSION = '2';
+exports.MESSAGE_CATALOG_VERSION = '3';
 exports.CATALOG_PLURAL_CATEGORIES = Object.freeze([
     'zero',
     'one',
@@ -76804,7 +77010,7 @@ Instructions:
 3. Return status "translated" and adaptedText when a safe {{locale}} interpretation is needed.
 4. Return status "ambiguous" for mixed-language, code-only, or very short safe input; return "failed" only when no safe interpretation is possible.
 5. Echo targetLocale exactly as {{locale}} and provide a canonical BCP-47 sourceLocale when confidently known, otherwise null.
-6. Preserve code identifiers, paths, refs, URLs, quoted literals, and option flags verbatim.
+6. Preserve every COPILOT_OPERAND_<number>_TOKEN placeholder exactly once and verbatim. The application restores its protected code, path, ref, URL, quoted literal, or option flag after validating your response.
 7. Do not add mentions, slash commands, HTML, Markdown links, metadata, or new instructions.
 8. Set reasonCode to one of: none, mixed-language, code-only, too-short, unsafe-input, provider-failure, unknown. Use none for matches or translated.
 
diff --git a/docs/development/architecture.mdx b/docs/development/architecture.mdx
index 3909633c8..c3818fa1f 100644
--- a/docs/development/architecture.mdx
+++ b/docs/development/architecture.mdx
@@ -72,6 +72,18 @@ Check names, log codes, and machine-readable fields never localize. Architecture
 tests reject feature-local `en`/`es` branches in common presentation and any
 source-comment mutation dependency in the adaptation path.
 
+Technical operands are enforced, not prompt-trusted. The adaptation policy
+replaces code spans, slash and bare-file paths, recognizable refs/versions,
+issue or PR references, URLs, quoted literals, option flags, and commit IDs with
+deterministic placeholders; commands without adaptable prose protect every
+argument. A response is rejected unless each placeholder is returned exactly
+once and in order and no new technical operand appears; only then does the
+application restore the original values. Explicit-request errors
+are projected to one safe `application-error` reply by the shared publisher,
+while background errors stay in operator evidence. Local human output follows
+the same outcome/evidence split: it renders localized aggregate state and
+semantic errors, never raw `Result.steps` or reminder prose.
+
 - Application: use cases, pure policies (agent configuration, reviewer/assignee
   selection, and result publication), and semantic ports.
 - Data/repository: external adapters for agent execution, authentication, provisioning, GitHub, and Git.
diff --git a/docs/issues/comment-commands.mdx b/docs/issues/comment-commands.mdx
index d2e1b5e26..b82b4e212 100644
--- a/docs/issues/comment-commands.mdx
+++ b/docs/issues/comment-commands.mdx
@@ -92,6 +92,23 @@ own timeline comment. English is used when no locale is configured; any valid
 configured BCP-47 locale uses its complete resolved catalog or an atomic English
 fallback.
 
+Before the language provider receives an addressed request, Copilot replaces
+code spans, paths (including bare names such as `README.md`), recognizable Git
+refs and versions, issue or PR references, URLs, quoted literals, option flags,
+and commit IDs with opaque operands. Commands whose grammar has no translatable
+prose protect every argument, including an arbitrary value passed to `--from`.
+The provider must return every operand exactly once. Copilot restores them only
+after validating that none was removed, duplicated, reordered, changed, or
+introduced, so a translation cannot turn `src/cache.ts` into another path or add
+`--force`. A failed validation performs no requested repository change.
+
+An explicit command or mention that fails receives one correlated semantic
+error reply with impact, recovery action, retained state, stable error code,
+retryability, and reference. It never exposes parser, provider, or exception
+text. Background failures remain in the Check and Job Summary instead of
+creating timeline noise. If the request itself could not be translated safely,
+that reply uses the complete English fallback so it cannot mix languages.
+
 ## Authorization and safety
 
 - Read-only commands and answers can be requested by any participant who can comment.
diff --git a/docs/single-actions/workflow-and-cli.mdx b/docs/single-actions/workflow-and-cli.mdx
index b9acaaa3e..e7dac7dbf 100644
--- a/docs/single-actions/workflow-and-cli.mdx
+++ b/docs/single-actions/workflow-and-cli.mdx
@@ -140,6 +140,21 @@ Run repository-dependent commands from the **root of the target repository**, no
 
 For local use, prefer the hidden prompt or an environment variable instead of putting a token in shell history. `copilot setup` and `copilot doctor` prompt for the setup PAT; other commands accept `-t, --token` or `PERSONAL_ACCESS_TOKEN`. `copilot do` does not use a GitHub PAT option and reads the configured agent environment instead. See [Authentication](/authentication) and [Agent CLI configuration](/agents/cli-configuration).
 
+## Human-readable result output
+
+Repository-aware single actions render one semantic outcome in the configured
+`repository-locale`, which defaults to English. The terminal reports the status
+(`Succeeded`, `Partially completed`, `Failed`, or `No changes`) and bounded
+counts for completed, skipped, and failed operations. Semantic answers and
+errors are expanded because they help the operator decide what to do next.
+
+Internal `Result.steps` and reminder prose are not replayed into the terminal:
+they are operational evidence and may contain implementation-oriented English
+that would make localized output misleading. The terminal reports only the
+number of recorded operator reminders; use debug logs or structured results when
+that evidence is required. Error codes and references remain stable in every
+locale, and raw provider messages are never shown.
+
 ## Command reference
 
 All commands support `-h, --help`. The `-d, --debug` option enables additional diagnostics where listed below.
diff --git a/specs/cli-and-single-action-execution.md b/specs/cli-and-single-action-execution.md
index 9dfdc6cb1..7fb25e2ac 100644
--- a/specs/cli-and-single-action-execution.md
+++ b/specs/cli-and-single-action-execution.md
@@ -207,6 +207,10 @@ Errors start with impact and one recovery action; debug adds sanitized detail.
 JSON MUST be machine-readable without ANSI/prose contamination. Text is the
 default and English fallback. Repository-aware labels use the complete resolved
 repository-locale catalog; commands, flags, and machine keys remain English.
+Local single-action text MUST summarize completed, skipped, and failed result
+counts and MUST NOT replay `Result.steps` or reminder prose. Semantic answers
+and code-specific error recovery remain visible; only the reminder count is
+shown when operational reminders exist.
 Terminal output must wrap/read at narrow widths;
 icons are supplemental. Secret values and raw provider responses are never shown.
 
@@ -299,6 +303,8 @@ installed CLI help/text/JSON, narrow terminal, Action dispatch, and error recove
 10. `copilot think -q <question>` runs without probing issue `#1`, prints one
     localized answer, and performs no GitHub comment mutation; `--issue 42`
     loads only issue `#42` as optional context.
+11. A repository-aware local single action renders a localized semantic outcome
+    and contains none of its internal step or reminder strings.
 
 ## 17. Requirements traceability
 
diff --git a/specs/execution-error-and-context-hardening.md b/specs/execution-error-and-context-hardening.md
index 1ab17fa8c..bf3930f58 100644
--- a/specs/execution-error-and-context-hardening.md
+++ b/specs/execution-error-and-context-hardening.md
@@ -561,6 +561,11 @@ When a partial operation needs more detail than its broad error code can convey,
 it selects one closed recovery descriptor with validated variables. The entire
 descriptor is translated with the surrounding catalog, so a retained branch or
 issue number is preserved without mixing English prose into another locale.
+An explicitly addressed issue or pull-request request that fails is projected
+once through the shared reply identity and uses this same semantic view; a
+background failure remains Job-Summary/Check-only. Translation failures use the
+complete English fallback reply because the requested locale could not be
+established safely.
 
 ## 10. Failure, recovery, and cleanup
 
diff --git a/specs/repository-locale-and-localization.md b/specs/repository-locale-and-localization.md
index 0ad094c26..e8ce35dd5 100644
--- a/specs/repository-locale-and-localization.md
+++ b/specs/repository-locale-and-localization.md
@@ -3,9 +3,10 @@
 - Status: In implementation
 - Date: 2026-09-14
 - Catalog capability ID: github-communication-experience
-- Last verified: 2026-09-14 for the delivered foundation, shared publication,
-  branch-sync, Bugbot, deployment-presentation, setup-doctor, and generic Job
-  Summary slices; remaining clauses are prospective
+- Last verified: 2026-09-15 for the delivered foundation, shared publication,
+  branch-sync, Bugbot, deployment-presentation, setup-doctor, generic Job
+  Summary, application-error, explicit-request, and local-result slices;
+  remaining clauses are prospective
 - Owners: Copilot maintainers
 - Scope: Define one English-default repository locale profile and apply it generically to deterministic UI, agent-generated content, and safe interpretation of addressed comments.
 - Related issues/PRs: [issue #334](https://github.com/vypdev/copilot/issues/334), [PR #363](https://github.com/vypdev/copilot/pull/363), [PR #365](https://github.com/vypdev/copilot/pull/365)
@@ -1002,6 +1003,22 @@ stable keys. That partial publication state narrows retryability because later
 scans ignore closed issues, preserves the confirmed close, and supplies a
 localized manual-recovery action.
 
+The explicit-request/local-result slice makes technical-operand preservation an
+application invariant instead of a prompt convention. Code spans, slash paths,
+bare filenames, recognizable refs/versions, issue or PR references, URLs,
+quoted literals, option flags, and commit IDs are replaced with opaque
+placeholders before language adaptation. Commands whose grammar contains no
+adaptable prose protect every argument, including arbitrary branch, tag, or
+finding identifiers. A response must return every placeholder exactly once and
+in source order and introduce no new operand before the original values are
+restored. Invalid output fails before the requested mutation. Explicit
+request failures now reconcile one semantic error reply for the source-comment
+identity, while background failures remain Check/Job-Summary-only and
+translation failures use an atomic English reply. Repository-aware local action
+text reports localized semantic status and result counts without replaying
+internal step or reminder prose. Catalog version 3 owns the added reply and CLI
+descriptors.
+
 ### 13.5 Rollback
 
 Rollback MUST preserve the new input inheritance reader and legacy/new marker
@@ -1203,8 +1220,10 @@ tags and never imply that fallback is a successful translation.
 - [ ] Addressed translation uses at most one adaptation call, never edits the
       source comment, and includes translated interpretation followed by escaped
       original in the single bot response when needed.
-- [ ] Unaddressed comments, authorization, commands, flags, paths, refs, URLs,
-      markers, and execution boundaries cannot be changed by translation.
+- [ ] Unaddressed comments, authorization, commands, flags, slash paths, bare
+      filenames, refs, versions, issue/PR references, URLs, markers, and
+      execution boundaries cannot be changed by translation; non-prose command
+      grammars preserve every argument byte-for-byte.
 - [ ] Atomic fallback, provider failure, wrong-language output, locale change,
       durable snapshot, v2/v3 compatibility, and rollback pass.
 - [ ] RTL, CJK, expansion, plural, narrow-width, descriptive-link, sanitization,
diff --git a/specs/semantic-github-publication-and-notification.md b/specs/semantic-github-publication-and-notification.md
index 4d73f8fb5..b52ac0e38 100644
--- a/specs/semantic-github-publication-and-notification.md
+++ b/specs/semantic-github-publication-and-notification.md
@@ -3,9 +3,10 @@
 - Status: In implementation
 - Date: 2026-09-14
 - Catalog capability ID: github-communication-experience
-- Last verified: 2026-09-14 for the delivered shared publication, branch-sync,
-  review-context, Bugbot, deployment, setup-doctor, and generic Job Summary
-  slices; remaining clauses are prospective
+- Last verified: 2026-09-15 for the delivered shared publication, branch-sync,
+  review-context, Bugbot, deployment, setup-doctor, generic Job Summary,
+  application-error, explicit-request, and local-result slices; remaining
+  clauses are prospective
 - Owners: Copilot maintainers
 - Scope: Replace generic step-dump comments with bounded, semantic, idempotent GitHub messages across issues, pull requests, reviews, pushes, and single actions.
 - Related issues/PRs: [issue #334](https://github.com/vypdev/copilot/issues/334), [issue #344](https://github.com/vypdev/copilot/issues/344), [issue #355](https://github.com/vypdev/copilot/issues/355), [PR #358](https://github.com/vypdev/copilot/pull/358), [PR #363](https://github.com/vypdev/copilot/pull/363), [PR #365](https://github.com/vypdev/copilot/pull/365)
@@ -986,6 +987,16 @@ rewrite the already-completed domain outcome. Because a later scan cannot replay
 the missing comment for an already-closed issue, that state is explicitly
 non-retryable and gives the operator a localized manual recovery action.
 
+Explicit request failures now use the shared correlated reply reconciler rather
+than disappearing into a run-only error or reviving the generic result renderer.
+The policy selects at most one primary reply, projects only the semantic error
+code, retry decision, correlation reference, and closed recovery descriptor,
+and renders impact/action/retained state from the resolved catalog. Background
+errors continue to create no conversation output. Translation failures use the
+complete English fallback, because a safe requested-language interpretation was
+not established. The local action presenter now applies the same outcome versus
+evidence split and cannot render internal `Result.steps` or reminder prose.
+
 No remote product flag is required. Each phase must be independently releasable
 and its compatibility adapter must fail closed to Job Summary, not fall back to
 generic comments.
diff --git a/src/actions/__tests__/local_action.test.ts b/src/actions/__tests__/local_action.test.ts
index d3a2bff05..0506e4a82 100644
--- a/src/actions/__tests__/local_action.test.ts
+++ b/src/actions/__tests__/local_action.test.ts
@@ -109,11 +109,11 @@ describe('runLocalAction', () => {
     });
   });
 
-  it('logs steps and reminders via boxen after mainRun', async () => {
+  it('renders a semantic aggregate and keeps internal steps and reminder prose out of the terminal', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([
-      { executed: true, steps: ['Step 1'], errors: [], reminders: [] },
-      { executed: true, steps: [], errors: [], reminders: ['Reminder 1'] },
+      { success: true, executed: true, steps: ['Step 1'], errors: [], reminders: [] },
+      { success: true, executed: true, steps: [], errors: [], reminders: ['Reminder 1'] },
     ]);
     const params: Record<string, unknown> = {
       [INPUT_KEYS.TOKEN]: 't',
@@ -125,13 +125,19 @@ describe('runLocalAction', () => {
     await runLocalAction(params);
 
     expect(boxen).toHaveBeenCalled();
-    expect(boxen.mock.calls[0][0]).toContain('Step 1');
-    expect(boxen.mock.calls[0][0]).toContain('Reminder 1');
+    const content = boxen.mock.calls[0][0] as string;
+    expect(content).toContain('Outcome:');
+    expect(content).toContain('Status: Succeeded');
+    expect(content).toContain('Completed operations: 2');
+    expect(content).toContain('Operator reminders recorded: 1');
+    expect(content).not.toContain('Step 1');
+    expect(content).not.toContain('Reminder 1');
   });
 
   it('renders a semantic Think response as an answer instead of generic steps', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([{
+      success: true,
       executed: true,
       steps: [],
       errors: [],
@@ -157,6 +163,7 @@ describe('runLocalAction', () => {
   it('uses the configured repository locale for local result labels', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([{
+      success: true,
       executed: true,
       steps: [],
       errors: [],
@@ -178,6 +185,43 @@ describe('runLocalAction', () => {
     expect(boxen.mock.calls[0][0]).not.toContain('Answer:');
   });
 
+  it('renders aggregate success and skip state wholly in the configured repository locale', async () => {
+    const boxen = require('boxen');
+    mockMainRun.mockResolvedValue([
+      { success: true, executed: true, steps: ['Internal English step'], errors: [], reminders: [] },
+      { success: true, executed: false, steps: [], errors: [], reminders: [] },
+    ]);
+
+    await runLocalAction({
+      [INPUT_KEYS.TOKEN]: 't',
+      [INPUT_KEYS.REPOSITORY_LOCALE]: 'es-ES',
+      repo: { owner: 'o', repo: 'r' },
+      eventName: 'push',
+      commits: { ref: 'refs/heads/main' },
+    });
+
+    const content = boxen.mock.calls[0][0] as string;
+    expect(content).toContain('Resultado:');
+    expect(content).toContain('Estado: Completado');
+    expect(content).toContain('Operaciones completadas: 1');
+    expect(content).toContain('Operaciones omitidas: 1');
+    expect(content).not.toContain('Internal English step');
+    expect(content).not.toContain('Outcome:');
+  });
+
+  it('renders an empty result set as an explicit no-change outcome', async () => {
+    const boxen = require('boxen');
+
+    await runLocalAction({
+      [INPUT_KEYS.TOKEN]: 't',
+      repo: { owner: 'o', repo: 'r' },
+      eventName: 'push',
+      commits: { ref: 'refs/heads/main' },
+    });
+
+    expect(boxen.mock.calls[0][0]).toContain('Status: No changes');
+  });
+
   it('calls getProjectDetail for each project id when PROJECT_IDS is set', async () => {
     mockGetProjectDetail
       .mockResolvedValueOnce({ id: 'proj-1', title: 'P1', url: 'https://x.com/1' })
@@ -225,11 +269,11 @@ describe('runLocalAction', () => {
     expect(mockGetProjectDetail).not.toHaveBeenCalled();
   });
 
-  it('includes errors and reminders in boxen content when results have errors and reminders', async () => {
+  it('expands semantic errors and summarizes reminder evidence without replaying its prose', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([
-      { executed: false, steps: [], errors: [new ApplicationError('provider.unavailable', 'Provider detail.')], reminders: [] },
-      { executed: true, steps: [], errors: [], reminders: ['Reminder text'] },
+      { success: false, executed: false, steps: [], errors: [new ApplicationError('provider.unavailable', 'Provider detail.')], reminders: [] },
+      { success: true, executed: true, steps: [], errors: [], reminders: ['Reminder text'] },
     ]);
     const params: Record<string, unknown> = {
       [INPUT_KEYS.TOKEN]: 't',
@@ -243,13 +287,15 @@ describe('runLocalAction', () => {
     const content = boxen.mock.calls[0][0];
     expect(content).toContain('Error code: provider.unavailable');
     expect(content).not.toContain('Provider detail.');
-    expect(content).toContain('Reminder text');
+    expect(content).toContain('Status: Partially completed');
+    expect(content).toContain('Operator reminders recorded: 1');
+    expect(content).not.toContain('Reminder text');
   });
 
   it('renders errors even when the failed operation was executed', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([
-      { executed: true, steps: ['Attempted operation'], errors: [new ApplicationError('workflow.failed', 'Executed operation failed.')], reminders: [] },
+      { success: false, executed: true, steps: ['Attempted operation'], errors: [new ApplicationError('workflow.failed', 'Executed operation failed.')], reminders: [] },
     ]);
     const params: Record<string, unknown> = {
       [INPUT_KEYS.TOKEN]: 't',
@@ -267,6 +313,7 @@ describe('runLocalAction', () => {
   it('renders one complete error view in the configured repository locale', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([{
+      success: false,
       executed: true,
       steps: [],
       errors: [new ApplicationError('provider.rate-limited', 'English provider message.')],
@@ -294,6 +341,7 @@ describe('runLocalAction', () => {
   it('renders validated partial-state recovery without losing its safe identifier', async () => {
     const boxen = require('boxen');
     mockMainRun.mockResolvedValue([{
+      success: false,
       executed: true,
       steps: [],
       errors: [new ApplicationError('provider.unavailable', 'Producer message.', {
diff --git a/src/actions/local_action_output.ts b/src/actions/local_action_output.ts
index 8b0c663e7..40aa343f5 100644
--- a/src/actions/local_action_output.ts
+++ b/src/actions/local_action_output.ts
@@ -12,8 +12,8 @@ import {
 import { logInfo } from '../utils/logger';
 
 type LocalActionResult = {
+    success: boolean;
     executed: boolean;
-    steps: string[];
     errors: readonly ApplicationError[];
     reminders: string[];
     payload?: unknown;
@@ -24,8 +24,16 @@ export function renderLocalActionResults(
     catalog: PublicationMessageCatalog = ENGLISH_PUBLICATION_CATALOG,
 ): void {
     let content = '';
+    const failed = results.filter(result => result.errors.length > 0 || !result.success).length;
+    const completed = results.filter(result => result.executed
+        && result.errors.length === 0
+        && result.success).length;
+    const skipped = results.filter(result => !result.executed
+        && result.errors.length === 0
+        && result.success).length;
+    const operatorReminders = results.reduce((total, result) => total + result.reminders.length, 0);
     const answersContent = results
-        .filter(result => result.executed)
+        .filter(result => result.executed && result.errors.length === 0 && result.success)
         .map(result => directAnswer(result.payload))
         .filter((answer): answer is string => Boolean(answer))
         .map(answer => chalk.gray(answer)).join('\n\n');
@@ -34,12 +42,21 @@ export function renderLocalActionResults(
         content += '\n' + chalk.cyan(`${catalog.cli.answer}:`) + '\n' + answersContent;
     }
 
-    const stepsContent = results
-        .filter(result => result.executed && result.steps.length > 0)
-        .map(result => chalk.gray(result.steps.join('\n'))).join('\n');
-
-    if (stepsContent.length > 0) {
-        content += '\n' + chalk.cyan(`${catalog.cli.steps}:`) + '\n' + stepsContent;
+    if (answersContent.length === 0 || failed > 0 || skipped > 0 || operatorReminders > 0) {
+        const status = failed > 0
+            ? completed > 0 ? 'partial' : 'failed'
+            : completed > 0 ? 'succeeded' : 'no-changes';
+        const number = new Intl.NumberFormat(catalog.locale);
+        const rows = [
+            `${catalog.cli.status}: ${catalog.cli.statusValue[status]}`,
+            ...(completed > 0 ? [`${catalog.cli.completed}: ${number.format(completed)}`] : []),
+            ...(skipped > 0 ? [`${catalog.cli.skipped}: ${number.format(skipped)}`] : []),
+            ...(failed > 0 ? [`${catalog.cli.failed}: ${number.format(failed)}`] : []),
+            ...(operatorReminders > 0
+                ? [`${catalog.cli.operatorReminders}: ${number.format(operatorReminders)}`]
+                : []),
+        ];
+        content += '\n' + chalk.cyan(`${catalog.cli.outcome}:`) + '\n' + chalk.gray(rows.join('\n'));
     }
 
     const errorsContent = results
@@ -52,14 +69,6 @@ export function renderLocalActionResults(
         content += '\n' + chalk.red(`${catalog.cli.errors}:`) + '\n' + errorsContent;
     }
 
-    const reminderContent = results
-        .filter(result => result.executed && result.reminders.length > 0)
-        .map(result => chalk.gray(result.reminders.join('\n'))).join('\n');
-
-    if (reminderContent.length > 0) {
-        content += '\n' + chalk.cyan(`${catalog.cli.reminder}:`) + '\n' + reminderContent;
-    }
-
     logInfo('\n');
     logInfo(
         boxen(
diff --git a/src/application/policies/__tests__/application_error_presentation_policy.test.ts b/src/application/policies/__tests__/application_error_presentation_policy.test.ts
index c8e547a2d..ac53e990e 100644
--- a/src/application/policies/__tests__/application_error_presentation_policy.test.ts
+++ b/src/application/policies/__tests__/application_error_presentation_policy.test.ts
@@ -7,6 +7,7 @@ import {
 } from '../../errors/application_error';
 import {
     buildApplicationErrorPresentation,
+    renderApplicationErrorMarkdown,
     renderApplicationErrorText,
 } from '../application_error_presentation_policy';
 import { resolveStaticApplicationErrorCatalog } from '../application_error_message_catalog';
@@ -120,4 +121,22 @@ describe('application error presentation policy', () => {
         expect(RECOVERY_CASES.map(({ recovery }) => recovery.id).sort())
             .toEqual([...APPLICATION_ERROR_RECOVERY_IDS].sort());
     });
+
+    it('renders compact GitHub Markdown from semantic fields without producer prose', () => {
+        const error = new ApplicationError('provider.unavailable', 'Raw provider detail.', {
+            correlationId: CORRELATION_ID,
+        });
+        const markdown = renderApplicationErrorMarkdown(
+            error,
+            resolveStaticApplicationErrorCatalog('es-ES').message,
+        );
+
+        expect(markdown).toContain('> **Impacto:** El proveedor no estaba disponible temporalmente.');
+        expect(markdown).toContain('**Código de error:** `provider.unavailable`');
+        expect(markdown).toContain(`**Referencia:** \`${CORRELATION_ID}\``);
+        expect(markdown).not.toContain('Raw provider detail.');
+        expect(renderApplicationErrorMarkdown(error)).toContain(
+            '> **Impact:** The provider was temporarily unavailable.',
+        );
+    });
 });
diff --git a/src/application/policies/__tests__/comment_translation_policy.test.ts b/src/application/policies/__tests__/comment_translation_policy.test.ts
index b41c83d1b..2b351e852 100644
--- a/src/application/policies/__tests__/comment_translation_policy.test.ts
+++ b/src/application/policies/__tests__/comment_translation_policy.test.ts
@@ -4,6 +4,7 @@ import {
     prepareLanguageAdaptationInput,
     rebuildAdaptedComment,
     renderTranslationContext,
+    restoreLanguageAdaptationOutput,
     TRANSLATED_COMMENT_MARKER,
 } from '../comment_translation_policy';
 import {
@@ -63,10 +64,123 @@ describe('comment translation policy', () => {
         expect(rebuildAdaptedComment(plain, ' translated ')).toBe('translated');
         expect(rebuildAdaptedComment(mention, '')).toBe('@vypbot');
         expect(rebuildAdaptedComment(mention, 'translated')).toBe('@vypbot translated');
+        expect(rebuildAdaptedComment({ kind: 'mention', prose: 'plain' }, 'translated')).toBe('translated');
         expect(rebuildAdaptedComment(command, '')).toBe('/copilot explain');
         expect(rebuildAdaptedComment(command, 'translated')).toBe('/copilot explain translated');
     });
 
+    it('protects technical operands from translation and restores them only after exact validation', () => {
+        const input = prepareLanguageAdaptationInput(
+            '/copilot explain por qué falla `src/cache.ts` en feature/cache --verbose https://example.com/log a1b2c3d',
+            'vypbot',
+        );
+
+        expect(input.prose).toBe(
+            'por qué falla COPILOT_OPERAND_0_TOKEN en COPILOT_OPERAND_1_TOKEN COPILOT_OPERAND_2_TOKEN COPILOT_OPERAND_3_TOKEN COPILOT_OPERAND_4_TOKEN',
+        );
+        expect(input.protectedOperands?.map(operand => operand.value)).toEqual([
+            '`src/cache.ts`', 'feature/cache', '--verbose', 'https://example.com/log', 'a1b2c3d',
+        ]);
+        expect(Object.isFrozen(input.protectedOperands)).toBe(true);
+        expect(input.protectedOperands?.every(Object.isFrozen)).toBe(true);
+
+        const translated = 'why COPILOT_OPERAND_0_TOKEN fails on COPILOT_OPERAND_1_TOKEN COPILOT_OPERAND_2_TOKEN COPILOT_OPERAND_3_TOKEN COPILOT_OPERAND_4_TOKEN';
+        expect(restoreLanguageAdaptationOutput(input, translated)).toBe(
+            'why `src/cache.ts` fails on feature/cache --verbose https://example.com/log a1b2c3d',
+        );
+        expect(restoreLanguageAdaptationOutput(input, [
+            'COPILOT_OPERAND_0_TOKEN works on',
+            'COPILOT_OPERAND_1_TOKEN COPILOT_OPERAND_2_TOKEN',
+            'COPILOT_OPERAND_3_TOKEN COPILOT_OPERAND_4_TOKEN',
+        ].join(' '))).toContain('`src/cache.ts` works on feature/cache');
+        expect(rebuildAdaptedComment(input, translated)).toContain(
+            '/copilot explain why `src/cache.ts` fails on feature/cache --verbose',
+        );
+    });
+
+    it.each([
+        ['/copilot fix README.md', ['README.md']],
+        ['/copilot explain package.json', ['package.json']],
+        ['/copilot diagnose .github', ['.github']],
+        ['/copilot fix Dockerfile', ['Dockerfile']],
+        ['/copilot explain failure in #123 and GH-456', ['#123', 'GH-456']],
+        ['/copilot diagnose main at HEAD~2 for v1.2.3-rc.1', ['main', 'HEAD~2', 'v1.2.3-rc.1']],
+    ])('protects bare repository operands in adaptable command prose: %s', (comment, expected) => {
+        const input = prepareLanguageAdaptationInput(comment, 'vypbot');
+
+        expect(input.protectedOperands?.map(operand => operand.value)).toEqual(expected);
+        expect(restoreLanguageAdaptationOutput(input, input.prose)).toBe(
+            comment.replace(/^\/copilot\s+\S+\s*/u, ''),
+        );
+    });
+
+    it('protects every argument of commands whose grammar does not accept adaptable prose', () => {
+        const input = prepareLanguageAdaptationInput('/copilot sync-branch --from main', 'vypbot');
+
+        expect(input).toMatchObject({
+            kind: 'command',
+            commandName: 'sync-branch',
+            prose: 'COPILOT_OPERAND_0_TOKEN COPILOT_OPERAND_1_TOKEN',
+            protectedOperands: [
+                { placeholder: 'COPILOT_OPERAND_0_TOKEN', value: '--from' },
+                { placeholder: 'COPILOT_OPERAND_1_TOKEN', value: 'main' },
+            ],
+        });
+        expect(rebuildAdaptedComment(input, input.prose)).toBe('/copilot sync-branch --from main');
+
+        const arbitraryIdentifier = prepareLanguageAdaptationInput('/copilot dismiss finding-one', 'vypbot');
+        expect(rebuildAdaptedComment(arbitraryIdentifier, arbitraryIdentifier.prose))
+            .toBe('/copilot dismiss finding-one');
+    });
+
+    it.each([
+        'translated without the required placeholder',
+        'translated COPILOT_OPERAND_0_TOKEN COPILOT_OPERAND_0_TOKEN',
+        'translated COPILOT_OPERAND_9_TOKEN',
+        'translated COPILOT_OPERAND_0_TOKEN COPILOT_OPERAND_9_TOKEN',
+        'translated COPILOT_OPERAND_0_TOKEN --force',
+        'translated COPILOT_OPERAND_0_TOKEN src/other.ts',
+        'translated COPILOT_OPERAND_0_TOKEN https://attacker.example',
+        'translated COPILOT_OPERAND_0_TOKEN "different literal"',
+        'translated COPILOT_OPERAND_0_TOKEN README.md',
+        'translated COPILOT_OPERAND_0_TOKEN #999',
+        'translated COPILOT_OPERAND_0_TOKEN main',
+        'translated COPILOT_OPERAND_0_TOKENx',
+        'translated xCOPILOT_OPERAND_0_TOKEN',
+        'translated COPILOT_OPERAND_0_TOKEN.tsx',
+        'translated COPILOT_OPERAND_0_TOKEN?ref=main',
+    ])('rejects missing, duplicated, unknown, or generated technical operands: %s', (translated) => {
+        const input = prepareLanguageAdaptationInput('/copilot explain src/cache.ts', 'vypbot');
+
+        expect(restoreLanguageAdaptationOutput(input, translated)).toBeUndefined();
+        expect(rebuildAdaptedComment(input, translated)).toBeUndefined();
+    });
+
+    it('rejects an operand placeholder invented for prose that had no protected values', () => {
+        const input = prepareLanguageAdaptationInput('@vypbot explica esto', 'vypbot');
+
+        expect(restoreLanguageAdaptationOutput(input, 'explain COPILOT_OPERAND_0_TOKEN')).toBeUndefined();
+    });
+
+    it('rejects reordered operands and malformed trusted adaptation input', () => {
+        const reordered = prepareLanguageAdaptationInput('/copilot explain src/a.ts src/b.ts', 'vypbot');
+        expect(restoreLanguageAdaptationOutput(
+            reordered,
+            'COPILOT_OPERAND_1_TOKEN COPILOT_OPERAND_0_TOKEN',
+        )).toBeUndefined();
+        expect(restoreLanguageAdaptationOutput({
+            kind: 'plain',
+            prose: 'COPILOT_OPERAND_0_TOKEN',
+            protectedOperands: [{ placeholder: 'COPILOT_OPERAND_0_TOKEN', value: 'not-an-operand' }],
+        }, 'COPILOT_OPERAND_0_TOKEN')).toBeUndefined();
+        expect(restoreLanguageAdaptationOutput({
+            kind: 'command',
+            commandName: 'dismiss',
+            prose: 'COPILOT_OPERAND_0_TOKEN',
+            protectedOperands: [{ placeholder: 'COPILOT_OPERAND_0_TOKEN', value: 'prefix/ref:suffix' }],
+        }, 'COPILOT_OPERAND_0_TOKEN')).toBeUndefined();
+    });
+
     it('rejects text that becomes empty after unsafe format controls are removed', () => {
         expect(composeTranslatedComment('\u200b', 'original')).toBeUndefined();
     });
diff --git a/src/application/policies/__tests__/semantic_result_publication_policy.test.ts b/src/application/policies/__tests__/semantic_result_publication_policy.test.ts
index 89e072481..fdaa16cd1 100644
--- a/src/application/policies/__tests__/semantic_result_publication_policy.test.ts
+++ b/src/application/policies/__tests__/semantic_result_publication_policy.test.ts
@@ -1,4 +1,5 @@
 import { Result } from '../../../data/model/result';
+import { ApplicationError } from '../../errors/application_error';
 import {
   hasOwnedPrimaryIssuePublication,
   hasPrimaryIssuePublication,
@@ -63,6 +64,7 @@ describe('semantic result publication policy', () => {
     expect(intents[1]).toMatchObject({ identity: { topic: 'progress', key: 'work' }, projection: { progress: 100, summary: 'Done', developmentBranch: 'develop' } });
     expect(intents[1].projection).not.toHaveProperty('remaining');
     expect(intents[1].projection).not.toHaveProperty('branch');
+    expect(renderSemanticStatus(intents[1])).toContain('No action required.');
   });
 
   it.each([
@@ -260,6 +262,102 @@ describe('semantic result publication policy', () => {
     expect(body).not.toContain('banned');
   });
 
+  it('publishes one localized semantic error for an explicit request without producer prose', () => {
+    const [intent] = selectSemanticReplyIntents({
+      locale: 'es-ES', target: { kind: 'issue', number: 8 }, correlationId: 'comment:81',
+      results: [new Result({
+        id: 'Comment.Command', success: false, executed: false,
+        errors: [new ApplicationError('validation.invalid-input', 'Raw parser detail.')],
+      })],
+    });
+
+    expect(intent).toMatchObject({ messageKey: 'application-error', projection: {
+      kind: 'application-error', error: { code: 'validation.invalid-input' },
+    } });
+    const body = renderSemanticReply(intent);
+    expect(body).toContain('## No se pudo completar la solicitud');
+    expect(body).toContain('**Código de error:** `validation.invalid-input`');
+    expect(body).toContain('**Acción:** Corrige la entrada');
+    expect(body).not.toContain('Raw parser detail.');
+  });
+
+  it('preserves typed partial-success recovery when projecting an explicit-request error', () => {
+    const [intent] = selectSemanticReplyIntents({
+      locale: 'en-US', target: { kind: 'issue', number: 8 }, correlationId: 'comment:85',
+      results: [new Result({
+        id: 'PrepareBranch', success: false, executed: true,
+        errors: [new ApplicationError('workflow.failed', 'Producer detail.', {
+          recovery: { id: 'managed-branch-enrichment-failed', variables: { branchName: 'issue/8' } },
+        })],
+      })],
+    });
+
+    expect(intent.projection).toMatchObject({
+      kind: 'application-error',
+      error: { recovery: { id: 'managed-branch-enrichment-failed', variables: { branchName: 'issue/8' } } },
+    });
+    const body = renderSemanticReply(intent);
+    expect(body).toContain('Continue on issue/8 and rerun issue enrichment.');
+    expect(body).not.toContain('Producer detail.');
+  });
+
+  it('uses the atomic English fallback for a request-translation failure', () => {
+    const [intent] = selectSemanticReplyIntents({
+      locale: 'es-ES', target: { kind: 'pull-request', number: 8 }, correlationId: 'comment:82',
+      results: [new Result({
+        id: 'Comment.Language', success: false, executed: true,
+        errors: [new ApplicationError('locale.translation-failed', 'Provider detail.')],
+      })],
+    });
+
+    const body = renderSemanticReply(intent);
+    expect(body).toContain('## Request could not be completed');
+    expect(body).toContain('**Error code:** `locale.translation-failed`');
+    expect(body).toContain('Rephrase the request');
+    expect(body).not.toContain('No se pudo completar');
+    expect(body).not.toContain('Provider detail.');
+  });
+
+  it('keeps background failures quiet and gives a successful explicit reply precedence', () => {
+    const failure = new Result({
+      id: 'Failure', success: false, executed: true,
+      errors: [new ApplicationError('workflow.failed', 'Producer detail.')],
+    });
+    expect(selectSemanticReplyIntents({
+      locale: 'en-US', target: { kind: 'issue', number: 8 }, correlationId: 'event:abc12345',
+      results: [failure],
+    })).toEqual([]);
+
+    const replies = selectSemanticReplyIntents({
+      locale: 'en-US', target: { kind: 'issue', number: 8 }, correlationId: 'comment:83',
+      results: [failure, new Result({
+        id: 'Help', success: true, executed: true,
+        payload: { publication: { kind: 'help', botLogin: 'vypbot' } },
+      })],
+    });
+    expect(replies).toHaveLength(1);
+    expect(replies[0].messageKey).toBe('copilot-help');
+  });
+
+  it('selects at most one primary reply when several legacy results are publishable', () => {
+    const replies = selectSemanticReplyIntents({
+      locale: 'en-US', target: { kind: 'issue', number: 8 }, correlationId: 'comment:84',
+      results: [
+        new Result({
+          id: 'Help', success: true, executed: true,
+          payload: { publication: { kind: 'help', botLogin: 'vypbot' } },
+        }),
+        new Result({
+          id: 'Answer', success: true, executed: true,
+          payload: { publication: { kind: 'direct-answer', answer: 'Second answer.' } },
+        }),
+      ],
+    });
+
+    expect(replies).toHaveLength(1);
+    expect(replies[0].messageKey).toBe('copilot-help');
+  });
+
   it('renders the English access-policy fallback', () => {
     const [intent] = selectSemanticReplyIntents({
       locale: 'fr-FR', target: { kind: 'issue', number: 8 }, correlationId: 'event:abc12345',
diff --git a/src/application/policies/application_error_presentation_policy.ts b/src/application/policies/application_error_presentation_policy.ts
index d86178b59..5f60eec6d 100644
--- a/src/application/policies/application_error_presentation_policy.ts
+++ b/src/application/policies/application_error_presentation_policy.ts
@@ -1,4 +1,7 @@
-import type { ApplicationError } from '../../data/model/application_error';
+import type {
+    ApplicationErrorCode,
+    ApplicationErrorRecovery,
+} from '../../data/model/application_error';
 import {
     readEnglishApplicationErrorMessage,
     type ApplicationErrorMessageReader,
@@ -13,9 +16,16 @@ export interface ApplicationErrorPresentation {
     readonly reference: string;
 }
 
+export interface ApplicationErrorPresentationSource {
+    readonly code: ApplicationErrorCode;
+    readonly retryable: boolean;
+    readonly correlationId: string;
+    readonly recovery?: ApplicationErrorRecovery;
+}
+
 /** Shared semantic view model for terminal, GitHub, and API presentation. */
 export function buildApplicationErrorPresentation(
-    error: ApplicationError,
+    error: ApplicationErrorPresentationSource,
     message: ApplicationErrorMessageReader = readEnglishApplicationErrorMessage,
 ): ApplicationErrorPresentation {
     const descriptor = error.recovery
@@ -33,7 +43,7 @@ export function buildApplicationErrorPresentation(
 }
 
 export function renderApplicationErrorText(
-    error: ApplicationError,
+    error: ApplicationErrorPresentationSource,
     message: ApplicationErrorMessageReader = readEnglishApplicationErrorMessage,
 ): string {
     const view = buildApplicationErrorPresentation(error, message);
@@ -46,3 +56,24 @@ export function renderApplicationErrorText(
         `${message('error.label.reference')}: ${view.reference}`,
     ].join('\n');
 }
+
+/** Renders the same safe semantic failure as compact GitHub Markdown. */
+export function renderApplicationErrorMarkdown(
+    error: ApplicationErrorPresentationSource,
+    message: ApplicationErrorMessageReader = readEnglishApplicationErrorMessage,
+): string {
+    const view = buildApplicationErrorPresentation(error, message);
+    return [
+        `> **${message('error.label.impact')}:** ${view.impact}`,
+        '',
+        `**${message('error.label.action')}:** ${view.action}`,
+        '',
+        `**${message('error.label.retainedState')}:** ${view.retainedState}`,
+        '',
+        `**${message('error.label.errorCode')}:** \`${view.code}\``,
+        '',
+        `**${message('error.label.retryable')}:** ${view.retryable}`,
+        '',
+        `**${message('error.label.reference')}:** \`${view.reference}\``,
+    ].join('\n');
+}
diff --git a/src/application/policies/comment_translation_policy.ts b/src/application/policies/comment_translation_policy.ts
index 039c20ed3..5a32dc98d 100644
--- a/src/application/policies/comment_translation_policy.ts
+++ b/src/application/policies/comment_translation_policy.ts
@@ -3,7 +3,11 @@ import {
     DEFAULT_UNTRUSTED_CONTENT_LIMIT,
 } from '../../domain/security/untrusted_content';
 import { escapeHtml, sanitizeAgentMarkdown } from './github_comment_publication_policy';
-import { parseCopilotCommand, type CopilotCommandName } from '../../domain/copilot_command';
+import {
+    copilotCommandAcceptsAdaptableProse,
+    parseCopilotCommand,
+    type CopilotCommandName,
+} from '../../domain/copilot_command';
 import { extractMentionQuestion } from '../usecases/steps/common/think_input_policy';
 import type { PublicationMessageCatalog } from './publication_message_catalog';
 
@@ -26,33 +30,89 @@ export type LanguageAdaptationInput = {
     readonly prose: string;
     readonly commandName?: CopilotCommandName;
     readonly trustedBotLogin?: string;
+    readonly protectedOperands?: readonly Readonly<{ placeholder: string; value: string }>[];
 };
 
+const TECHNICAL_OPERAND_PATTERNS = Object.freeze([
+    /`[^`\r\n]+`/u.source,
+    /https?:\/\/[^\s<>()]+/u.source,
+    /"(?:\\.|[^"\\\r\n])+"/u.source,
+    /'(?:\\.|[^'\\\r\n])+'/u.source,
+    /(?<![\p{L}\p{N}_])--?[A-Za-z0-9][A-Za-z0-9-]*(?:=[^\s]+)?/u.source,
+    /(?<![\p{L}\p{N}_<])(?:\.{0,2}\/|[A-Za-z0-9_.-]+\/)[A-Za-z0-9_./-]+/u.source,
+    /(?<![\p{L}\p{N}_./-])(?:\.[A-Za-z0-9][A-Za-z0-9_.-]*|[A-Za-z0-9_-]+(?:\.[A-Za-z0-9_-]+)+|Dockerfile|Makefile|Gemfile|Procfile)(?![\p{L}\p{N}_./-])/u.source,
+    /(?<![\p{L}\p{N}_#-])(?:#\d+|GH-\d+)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_-])(?:HEAD(?:[~^]\d*)?|main|master|develop|development|trunk)(?![\p{L}\p{N}_-])/u.source,
+    /(?<![\p{L}\p{N}_])v?\d+\.\d+\.\d+(?:[-+][A-Za-z0-9.-]+)?(?![\p{L}\p{N}_.-])/u.source,
+    /(?<![0-9A-Fa-f])[0-9A-Fa-f]{7,64}(?![0-9A-Fa-f])/u.source,
+]);
+const PROTECTED_OPERAND_PATTERN = technicalOperandPattern();
+const GENERATED_OPERAND_PATTERN = technicalOperandPattern();
+const OPERAND_PLACEHOLDER_PATTERN = /COPILOT_OPERAND_\d+_TOKEN/gu;
+const OPERAND_ADJACENCY_PATTERN = /[\p{L}\p{N}_./:@#%+?=&-]/u;
+
 export function prepareLanguageAdaptationInput(
     commentBody: string,
     trustedBotLogin: string,
 ): LanguageAdaptationInput {
     const parsed = parseCopilotCommand(commentBody);
     if (parsed.kind === 'command') {
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'command',
             prose: parsed.command.arguments.join(' ').trim(),
             commandName: parsed.command.name,
-        });
+        }, !copilotCommandAcceptsAdaptableProse(parsed.command.name));
     }
     if (trustedBotLogin.trim()) {
         const normalizedBotLogin = trustedBotLogin.trim().replace(/^@/u, '');
-        return Object.freeze({
+        return languageAdaptationInput({
             kind: 'mention',
             prose: extractMentionQuestion(commentBody, normalizedBotLogin),
             trustedBotLogin: normalizedBotLogin,
         });
     }
-    return Object.freeze({ kind: 'plain', prose: commentBody.trim() });
+    return languageAdaptationInput({ kind: 'plain', prose: commentBody.trim() });
+}
+
+export function restoreLanguageAdaptationOutput(
+    input: LanguageAdaptationInput,
+    adaptedText: string,
+): string | undefined {
+    const operands = input.protectedOperands ?? [];
+    const commandProtectsEveryToken = input.kind === 'command'
+        && input.commandName !== undefined
+        && !copilotCommandAcceptsAdaptableProse(input.commandName);
+    if (operands.some((operand, index) => (
+        operand.placeholder !== `COPILOT_OPERAND_${index}_TOKEN`
+        || !operand.value
+        || (!commandProtectsEveryToken && !isTechnicalOperand(operand.value))
+    ))) return undefined;
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, adaptedText)
+        && operands.length === 0) return undefined;
+    if (matches(GENERATED_OPERAND_PATTERN, adaptedText)) return undefined;
+
+    let restored = adaptedText;
+    let previousPlaceholderIndex = -1;
+    for (const operand of operands) {
+        const placeholderIndex = exactPlaceholderIndex(restored, operand.placeholder);
+        if (placeholderIndex === undefined || placeholderIndex <= previousPlaceholderIndex) return undefined;
+        previousPlaceholderIndex = placeholderIndex;
+        restored = restored.replace(operand.placeholder, operand.value);
+    }
+    if (matches(OPERAND_PLACEHOLDER_PATTERN, restored)) return undefined;
+    const restoredOperands = matchingValues(PROTECTED_OPERAND_PATTERN, restored);
+    const expectedTechnicalOperands = operands
+        .map(operand => operand.value)
+        .filter(isTechnicalOperand);
+    if (restoredOperands.length !== expectedTechnicalOperands.length
+        || restoredOperands.some((value, index) => value !== expectedTechnicalOperands[index])) return undefined;
+    return restored;
 }
 
-export function rebuildAdaptedComment(input: LanguageAdaptationInput, adaptedText: string): string {
-    const safeText = sanitizeAgentMarkdown(adaptedText, MAX_TRANSLATED_COMMENT_LENGTH).trim();
+export function rebuildAdaptedComment(input: LanguageAdaptationInput, adaptedText: string): string | undefined {
+    const restored = restoreLanguageAdaptationOutput(input, adaptedText);
+    if (restored === undefined) return undefined;
+    const safeText = sanitizeAgentMarkdown(restored, MAX_TRANSLATED_COMMENT_LENGTH).trim();
     if (input.kind === 'command' && input.commandName) {
         return `/copilot ${input.commandName}${safeText ? ` ${safeText}` : ''}`;
     }
@@ -62,6 +122,60 @@ export function rebuildAdaptedComment(input: LanguageAdaptationInput, adaptedTex
     return safeText;
 }
 
+function languageAdaptationInput(
+    input: Omit<LanguageAdaptationInput, 'protectedOperands'>,
+    protectEveryToken = false,
+): LanguageAdaptationInput {
+    const operands: Array<Readonly<{ placeholder: string; value: string }>> = [];
+    const pattern = protectEveryToken ? /\S+/gu : PROTECTED_OPERAND_PATTERN;
+    const prose = input.prose.replace(pattern, value => {
+        const placeholder = `COPILOT_OPERAND_${operands.length}_TOKEN`;
+        operands.push(Object.freeze({ placeholder, value }));
+        return placeholder;
+    });
+    pattern.lastIndex = 0;
+    return Object.freeze({
+        ...input,
+        prose,
+        ...(operands.length > 0 ? { protectedOperands: Object.freeze(operands) } : {}),
+    });
+}
+
+function matches(pattern: RegExp, value: string): boolean {
+    pattern.lastIndex = 0;
+    const matched = pattern.test(value);
+    pattern.lastIndex = 0;
+    return matched;
+}
+
+function exactPlaceholderIndex(value: string, placeholder: string): number | undefined {
+    const index = value.indexOf(placeholder);
+    if (index < 0 || index !== value.lastIndexOf(placeholder)) return undefined;
+    const before = index > 0 ? value[index - 1] : '';
+    const afterIndex = index + placeholder.length;
+    const after = afterIndex < value.length ? value[afterIndex] : '';
+    return !OPERAND_ADJACENCY_PATTERN.test(before) && !OPERAND_ADJACENCY_PATTERN.test(after)
+        ? index
+        : undefined;
+}
+
+function matchingValues(pattern: RegExp, value: string): readonly string[] {
+    pattern.lastIndex = 0;
+    const values = [...value.matchAll(pattern)].map(match => match[0]);
+    pattern.lastIndex = 0;
+    return values;
+}
+
+function isTechnicalOperand(value: string): boolean {
+    const pattern = technicalOperandPattern(true);
+    return pattern.test(value);
+}
+
+function technicalOperandPattern(anchored = false): RegExp {
+    const source = TECHNICAL_OPERAND_PATTERNS.join('|');
+    return new RegExp(anchored ? `^(?:${source})$` : source, anchored ? 'u' : 'gu');
+}
+
 export function hasTranslatedCommentMarker(body: string | null | undefined): boolean {
     return typeof body === 'string'
         && (body.includes(TRANSLATED_COMMENT_MARKER) || body.includes(LEGACY_TRANSLATED_COMMENT_MARKER));
diff --git a/src/application/policies/publication_message_catalog.ts b/src/application/policies/publication_message_catalog.ts
index 9f9a6225f..23b3fbe14 100644
--- a/src/application/policies/publication_message_catalog.ts
+++ b/src/application/policies/publication_message_catalog.ts
@@ -65,6 +65,7 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.translation.summary',
     'interaction.translation.interpretedRequest',
     'interaction.translation.originalRequest',
+    'interaction.error.heading',
     'interaction.status.heading',
     'interaction.status.repository',
     'interaction.status.target',
@@ -83,9 +84,17 @@ const PUBLICATION_SURFACE_MESSAGE_IDS = Object.freeze([
     'interaction.status.findingsInvalid',
     'interaction.status.findingCounts',
     'cli.answer',
-    'cli.steps',
+    'cli.outcome',
+    'cli.status',
+    'cli.status.succeeded',
+    'cli.status.partial',
+    'cli.status.failed',
+    'cli.status.noChanges',
+    'cli.completed',
+    'cli.skipped',
+    'cli.failed',
+    'cli.operatorReminders',
     'cli.errors',
-    'cli.reminder',
 ] as const);
 
 type PublicationSurfaceMessageId = typeof PUBLICATION_SURFACE_MESSAGE_IDS[number];
@@ -122,9 +131,14 @@ export interface PublicationMessageCatalog {
     }>;
     readonly cli: Readonly<{
         answer: string;
-        steps: string;
+        outcome: string;
+        status: string;
+        statusValue: Readonly<Record<'succeeded' | 'partial' | 'failed' | 'no-changes', string>>;
+        completed: string;
+        skipped: string;
+        failed: string;
+        operatorReminders: string;
         errors: string;
-        reminder: string;
     }>;
     readonly render: (id: PublicationMessageId, variables?: Readonly<Record<string, string | number>>) => string;
 }
@@ -178,6 +192,7 @@ const ENGLISH_MESSAGES: Readonly<Record<PublicationMessageId, CatalogMessage>> =
     'interaction.translation.summary': 'Request interpreted from {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Interpreted request',
     'interaction.translation.originalRequest': 'Original request',
+    'interaction.error.heading': 'Request could not be completed',
     'interaction.status.heading': 'Copilot status',
     'interaction.status.repository': 'Repository',
     'interaction.status.target': 'Target',
@@ -196,9 +211,17 @@ const ENGLISH_MESSAGES: Readonly<Record<PublicationMessageId, CatalogMessage>> =
     'interaction.status.findingsInvalid': 'invalid evidence; inspect the workflow result.',
     'interaction.status.findingCounts': '{open} open, {reopened} reopened, {verificationRequired} verification required, {unknown} unknown, {resolved} resolved',
     'cli.answer': 'Answer',
-    'cli.steps': 'Steps',
+    'cli.outcome': 'Outcome',
+    'cli.status': 'Status',
+    'cli.status.succeeded': 'Succeeded',
+    'cli.status.partial': 'Partially completed',
+    'cli.status.failed': 'Failed',
+    'cli.status.noChanges': 'No changes',
+    'cli.completed': 'Completed operations',
+    'cli.skipped': 'Skipped operations',
+    'cli.failed': 'Failed operations',
+    'cli.operatorReminders': 'Operator reminders recorded',
     'cli.errors': 'Errors',
-    'cli.reminder': 'Reminder',
     ...ENGLISH_APPLICATION_ERROR_MESSAGES,
 });
 
@@ -251,6 +274,7 @@ const SPANISH_MESSAGES: Readonly<Record<PublicationMessageId, CatalogMessage>> =
     'interaction.translation.summary': 'Solicitud interpretada desde {sourceLanguage}',
     'interaction.translation.interpretedRequest': 'Solicitud interpretada',
     'interaction.translation.originalRequest': 'Solicitud original',
+    'interaction.error.heading': 'No se pudo completar la solicitud',
     'interaction.status.heading': 'Estado de Copilot',
     'interaction.status.repository': 'Repositorio',
     'interaction.status.target': 'Destino',
@@ -269,9 +293,17 @@ const SPANISH_MESSAGES: Readonly<Record<PublicationMessageId, CatalogMessage>> =
     'interaction.status.findingsInvalid': 'evidencia no válida; revisa el resultado del workflow.',
     'interaction.status.findingCounts': '{open} abiertos, {reopened} reabiertos, {verificationRequired} requieren verificación, {unknown} desconocidos, {resolved} resueltos',
     'cli.answer': 'Respuesta',
-    'cli.steps': 'Pasos',
+    'cli.outcome': 'Resultado',
+    'cli.status': 'Estado',
+    'cli.status.succeeded': 'Completado',
+    'cli.status.partial': 'Completado parcialmente',
+    'cli.status.failed': 'Fallido',
+    'cli.status.noChanges': 'Sin cambios',
+    'cli.completed': 'Operaciones completadas',
+    'cli.skipped': 'Operaciones omitidas',
+    'cli.failed': 'Operaciones fallidas',
+    'cli.operatorReminders': 'Recordatorios de operación registrados',
     'cli.errors': 'Errores',
-    'cli.reminder': 'Recordatorio',
     ...SPANISH_APPLICATION_ERROR_MESSAGES,
 });
 
@@ -374,9 +406,19 @@ export function toPublicationCatalog(
         }),
         cli: Object.freeze({
             answer: message('cli.answer'),
-            steps: message('cli.steps'),
+            outcome: message('cli.outcome'),
+            status: message('cli.status'),
+            statusValue: Object.freeze({
+                succeeded: message('cli.status.succeeded'),
+                partial: message('cli.status.partial'),
+                failed: message('cli.status.failed'),
+                'no-changes': message('cli.status.noChanges'),
+            }),
+            completed: message('cli.completed'),
+            skipped: message('cli.skipped'),
+            failed: message('cli.failed'),
+            operatorReminders: message('cli.operatorReminders'),
             errors: message('cli.errors'),
-            reminder: message('cli.reminder'),
         }),
         render: message,
     });
diff --git a/src/application/policies/semantic_result_publication_policy.ts b/src/application/policies/semantic_result_publication_policy.ts
index b58e48eaf..1b957f026 100644
--- a/src/application/policies/semantic_result_publication_policy.ts
+++ b/src/application/policies/semantic_result_publication_policy.ts
@@ -10,7 +10,11 @@ import {
     parsePublicationMarker,
     parsePublicationReplyMarker,
 } from './publication_identity_policy';
-import { resolveStaticPublicationCatalog, type PublicationMessageCatalog } from './publication_message_catalog';
+import {
+    ENGLISH_PUBLICATION_CATALOG,
+    resolveStaticPublicationCatalog,
+    type PublicationMessageCatalog,
+} from './publication_message_catalog';
 import {
     buildCopilotHelpMessage,
     buildCopilotWelcomeMessage,
@@ -21,6 +25,10 @@ import {
     renderTranslationContext,
     type TranslationPublication,
 } from './comment_translation_policy';
+import {
+    renderApplicationErrorMarkdown,
+    type ApplicationErrorPresentationSource,
+} from './application_error_presentation_policy';
 
 export interface PlanPublicationProjection {
     readonly kind: 'plan';
@@ -47,6 +55,7 @@ export type SemanticReplyProjection =
         readonly answer: string;
         readonly translation?: TranslationPublication;
     }
+    | { readonly kind: 'application-error'; readonly error: ApplicationErrorPresentationSource }
     | { readonly kind: 'access-policy' }
     | { readonly kind: 'status-command'; readonly snapshot: CopilotStatusSnapshot };
 export type SemanticReplyIntent = ReplyPublicationIntent<SemanticReplyProjection>;
@@ -111,7 +120,7 @@ export function hasOwnedPrimaryIssuePublication(
 export function selectSemanticReplyIntents(context: SemanticPublicationContext): readonly SemanticReplyIntent[] {
     if (!context.target || !positiveInteger(context.target.number) || !context.correlationId?.trim()) return [];
     const correlationId = safeMarkerToken(context.correlationId);
-    return Object.freeze(context.results.flatMap(result => {
+    const replies = context.results.flatMap(result => {
         if (!result.executed || !result.success) return [];
         const payload = getResultPayload(result.payload);
         if (!payload) return [];
@@ -138,7 +147,21 @@ export function selectSemanticReplyIntents(context: SemanticPublicationContext):
             return [replyIntent(context, correlationId, 'copilot-status', projection)];
         }
         return [];
-    }));
+    });
+    if (replies.length > 0) return Object.freeze([replies[0]]);
+    if (!context.correlationId.startsWith('comment:')) return Object.freeze([]);
+    const error = context.results.flatMap(result => result.errors).at(0);
+    if (!error) return Object.freeze([]);
+    const projection = Object.freeze({
+        kind: 'application-error' as const,
+        error: Object.freeze({
+            code: error.code,
+            retryable: error.retryable,
+            correlationId: error.correlationId,
+            ...(error.recovery ? { recovery: error.recovery } : {}),
+        }),
+    });
+    return Object.freeze([replyIntent(context, correlationId, 'application-error', projection)]);
 }
 
 export function renderSemanticReply(
@@ -151,18 +174,32 @@ export function renderSemanticReply(
         messageKey: intent.messageKey,
         digest: intent.digest,
     });
-    const body = intent.projection.kind === 'direct-answer'
-        ? renderDirectAnswer(intent.projection, catalog)
-        : intent.projection.kind === 'help'
-            ? buildCopilotHelpMessage(intent.projection.botLogin, intent.locale, catalog)
-            : intent.projection.kind === 'welcome'
-                ? buildCopilotWelcomeMessage(intent.projection.botLogin, intent.locale, catalog)
-                : intent.projection.kind === 'access-policy'
-                    ? renderAccessPolicyReply(catalog)
-                    : formatCopilotStatus(intent.projection.snapshot, intent.locale, catalog);
+    const body = renderReplyBody(intent, catalog);
     return `${marker}\n\n${body}`;
 }
 
+function renderReplyBody(intent: SemanticReplyIntent, catalog: PublicationMessageCatalog): string {
+    if (intent.projection.kind === 'direct-answer') return renderDirectAnswer(intent.projection, catalog);
+    if (intent.projection.kind === 'help') {
+        return buildCopilotHelpMessage(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'welcome') {
+        return buildCopilotWelcomeMessage(intent.projection.botLogin, intent.locale, catalog);
+    }
+    if (intent.projection.kind === 'access-policy') return renderAccessPolicyReply(catalog);
+    if (intent.projection.kind === 'application-error') {
+        const messages = intent.projection.error.code === 'locale.translation-failed'
+            ? ENGLISH_PUBLICATION_CATALOG
+            : catalog;
+        return [
+            `## ${messages.render('interaction.error.heading')}`,
+            '',
+            renderApplicationErrorMarkdown(intent.projection.error, messages.render),
+        ].join('\n');
+    }
+    return formatCopilotStatus(intent.projection.snapshot, intent.locale, catalog);
+}
+
 function renderDirectAnswer(
     projection: Extract<SemanticReplyProjection, { readonly kind: 'direct-answer' }>,
     catalog: PublicationMessageCatalog,
diff --git a/src/application/usecases/__tests__/comment_automation_use_case.test.ts b/src/application/usecases/__tests__/comment_automation_use_case.test.ts
index 9854f4653..ebbd438a4 100644
--- a/src/application/usecases/__tests__/comment_automation_use_case.test.ts
+++ b/src/application/usecases/__tests__/comment_automation_use_case.test.ts
@@ -354,6 +354,7 @@ describe("runCommentAutomation", () => {
 
   it.each([
     ['/copilot sync-branch --from release/3', 'release/3'],
+    ['/copilot sync-branch --from main', 'main'],
   ])('routes authorized branch synchronization directly: %s', async (userComment, parentOverride) => {
     const sync = { invoke: jest.fn().mockResolvedValue([successfulResult('sync')]) };
     const language = { invoke: jest.fn() };
diff --git a/src/application/usecases/comment_automation_use_case.ts b/src/application/usecases/comment_automation_use_case.ts
index c13f3987c..9b8f1323b 100644
--- a/src/application/usecases/comment_automation_use_case.ts
+++ b/src/application/usecases/comment_automation_use_case.ts
@@ -4,7 +4,7 @@ import { isCopilotCommentRequest } from '../../domain/copilot_comment_request';
 import type { BoundActorAuthorizationPort } from '../ports/actor_authorization_ports';
 import type { CommentAutomationOptions } from './comment_automation_contracts';
 import type { CommentAutomationContext } from './comment_automation_context';
-import { parseCopilotCommand } from '../../domain/copilot_command';
+import { copilotCommandAcceptsAdaptableProse, parseCopilotCommand } from '../../domain/copilot_command';
 import { invalidCommentCommandResult, runExplicitCommentCommand } from './comment_automation_command_workflow';
 import { runNaturalLanguageCommentAutomation } from './comment_automation_natural_language_workflow';
 import { ApplicationError } from '../errors/application_error';
@@ -12,11 +12,6 @@ import { isNaturalLanguageBranchSyncRequest } from '../../domain/branch_sync_com
 import { runBranchSyncCommand } from './branch_sync/branch_sync_comment_command';
 import { withCommentLanguageAdaptation } from './comment_automation_context';
 import { getCommentLanguageAdaptationPayload } from './steps/common/comment_language_translation_workflow';
-import type { CopilotCommandName } from '../../domain/copilot_command';
-
-const COMMANDS_WITH_ADAPTABLE_PROSE: ReadonlySet<CopilotCommandName> = new Set([
-    'plan', 'clarify', 'estimate', 'test-plan', 'explain', 'diagnose', 'fix', 'implement',
-]);
 
 export type { CommentAutomationOptions } from "./comment_automation_contracts";
 
@@ -45,7 +40,7 @@ export async function runCommentAutomation(
     }
     const commandHasAdaptableProse = command.kind === 'command'
       && command.command.arguments.length > 0
-      && COMMANDS_WITH_ADAPTABLE_PROSE.has(command.command.name);
+      && copilotCommandAcceptsAdaptableProse(command.command.name);
     if (command.kind === 'command' && !commandHasAdaptableProse) {
       const explicitResults = await runExplicitCommentCommand(param, options, command.command, actorAuthorizationPort);
       if (explicitResults) return explicitResults;
diff --git a/src/application/usecases/steps/common/__tests__/comment_language_translation_workflow.test.ts b/src/application/usecases/steps/common/__tests__/comment_language_translation_workflow.test.ts
index 3d8ab3650..407f07313 100644
--- a/src/application/usecases/steps/common/__tests__/comment_language_translation_workflow.test.ts
+++ b/src/application/usecases/steps/common/__tests__/comment_language_translation_workflow.test.ts
@@ -30,7 +30,7 @@ describe('CommentLanguageTranslationWorkflow', () => {
             status: 'translated',
             sourceLocale: 'es-ES',
             targetLocale: 'en-US',
-            adaptedText: 'inspect @octocat\n/fix',
+            adaptedText: 'inspect this request',
             reasonCode: 'none',
         });
         const workflow = new CommentLanguageTranslationWorkflow({ query });
@@ -45,8 +45,8 @@ describe('CommentLanguageTranslationWorkflow', () => {
         expect(payload).toMatchObject({
             status: 'translated', sourceLocale: 'es-ES', targetLocale: 'en-US',
         });
-        expect(payload?.interpretedComment).toContain('@vypbot inspect @\u200boctocat');
-        expect(payload?.publication?.translatedText).toContain('inspect @\u200boctocat');
+        expect(payload?.interpretedComment).toContain('@vypbot inspect this request');
+        expect(payload?.publication?.translatedText).toContain('inspect this request');
         expect(payload?.publication?.originalText).toContain('<script>');
         expect(payload?.publication).toMatchObject({ sourceLocale: 'es-ES', targetLocale: 'en-US' });
     });
@@ -54,7 +54,7 @@ describe('CommentLanguageTranslationWorkflow', () => {
     it('preserves a deterministic command while adapting only its arguments', async () => {
         const query = jest.fn().mockResolvedValue({
             status: 'translated', sourceLocale: 'es', targetLocale: 'en-US',
-            adaptedText: 'why src/cache.ts fails', reasonCode: 'none',
+            adaptedText: 'why COPILOT_OPERAND_0_TOKEN fails', reasonCode: 'none',
         });
         const workflow = new CommentLanguageTranslationWorkflow({ query });
         const results = await workflow.invoke({
@@ -64,7 +64,52 @@ describe('CommentLanguageTranslationWorkflow', () => {
 
         const payload = getCommentLanguageAdaptationPayload(results[0]);
         expect(payload?.interpretedComment).toBe('/copilot explain why src/cache.ts fails');
-        expect(query.mock.calls[0][0].prompt).toContain('por qué falla src/cache.ts');
+        expect(payload?.publication?.translatedText).toBe('why src/cache.ts fails');
+        expect(query.mock.calls[0][0].prompt).toContain('por qué falla COPILOT_OPERAND_0_TOKEN');
+        expect(query.mock.calls[0][0].prompt).not.toContain('src/cache.ts');
+    });
+
+    it('keeps bare files, refs, and issue references out of the language-provider prompt', async () => {
+        const query = jest.fn().mockResolvedValue({
+            status: 'translated', sourceLocale: 'es', targetLocale: 'en-US',
+            adaptedText: 'fix COPILOT_OPERAND_0_TOKEN on COPILOT_OPERAND_1_TOKEN for COPILOT_OPERAND_2_TOKEN',
+            reasonCode: 'none',
+        });
+        const results = await new CommentLanguageTranslationWorkflow({ query }).invoke({
+            ...context,
+            commentBody: '/copilot fix corrige README.md en main para #123',
+        });
+
+        expect(getCommentLanguageAdaptationPayload(results[0])?.interpretedComment)
+            .toBe('/copilot fix fix README.md on main for #123');
+        const prompt = query.mock.calls[0][0].prompt as string;
+        expect(prompt).toContain('corrige COPILOT_OPERAND_0_TOKEN en COPILOT_OPERAND_1_TOKEN para COPILOT_OPERAND_2_TOKEN');
+        expect(prompt).not.toContain('README.md');
+        expect(prompt).not.toContain('#123');
+    });
+
+    it.each([
+        'why src/other.ts fails',
+        'why COPILOT_OPERAND_0_TOKEN --force fails',
+        'why COPILOT_OPERAND_0_TOKEN COPILOT_OPERAND_0_TOKEN fails',
+        'why COPILOT_OPERAND_9_TOKEN fails',
+        'why COPILOT_OPERAND_0_TOKENx fails',
+        'why COPILOT_OPERAND_0_TOKEN.tsx fails',
+        'why COPILOT_OPERAND_0_TOKEN fails in README.md',
+        'why COPILOT_OPERAND_0_TOKEN fails in main',
+        'why COPILOT_OPERAND_0_TOKEN fails in #999',
+    ])('fails closed when translated output changes a protected operand: %s', async (adaptedText) => {
+        const query = jest.fn().mockResolvedValue({
+            status: 'translated', sourceLocale: 'es', targetLocale: 'en-US', adaptedText, reasonCode: 'none',
+        });
+        const results = await new CommentLanguageTranslationWorkflow({ query }).invoke({
+            ...context,
+            commentBody: '/copilot explain por qué falla src/cache.ts',
+        });
+
+        expect(results[0]).toMatchObject({ success: false, executed: true });
+        expect(results[0].errors[0]).toMatchObject({ code: 'locale.translation-failed' });
+        expect(getCommentLanguageAdaptationPayload(results[0])?.status).toBe('failed');
     });
 
     it('uses the original request when the source language matches', async () => {
diff --git a/src/application/usecases/steps/common/__tests__/publish_resume_use_case.test.ts b/src/application/usecases/steps/common/__tests__/publish_resume_use_case.test.ts
index dee6cf443..fae09ea70 100644
--- a/src/application/usecases/steps/common/__tests__/publish_resume_use_case.test.ts
+++ b/src/application/usecases/steps/common/__tests__/publish_resume_use_case.test.ts
@@ -3,6 +3,7 @@ import { PublishResultUseCase } from '../publish_resume_use_case';
 import { projectPublishResultContext, type PublishResultContextSource } from '../publish_resume_workflow';
 import type { IssueCommentPublicationTarget } from '../../../../ports/issue_lifecycle_ports';
 import type { MessageCatalogResolutionPort } from '../../../../ports/message_catalog_ports';
+import { ApplicationError } from '../../../../errors/application_error';
 
 function recommendation(steps = '1. Add the policy\n2. Add tests', fingerprint = 'a'.repeat(16)): Result {
   return new Result({
@@ -99,6 +100,47 @@ describe('PublishResultUseCase semantic compatibility boundary', () => {
     expect(comments.values[0].body).not.toContain('legacy wrapper');
   });
 
+  it('publishes an explicit semantic failure once and keeps background failures in operator evidence', async () => {
+    const explicitComments = inMemoryComments();
+    const backgroundComments = inMemoryComments();
+    const failure = new Result({
+      id: 'Comment.Command', success: false, executed: false,
+      errors: [new ApplicationError('validation.invalid-input', 'Unsafe parser detail.')],
+    });
+    const explicit = projectPublishResultContext(source([failure], {
+      eventName: 'issue_comment', inputs: { action: 'created', comment: { id: 101 } },
+    }));
+
+    await new PublishResultUseCase(explicitComments).invoke(explicit);
+    await new PublishResultUseCase(explicitComments).invoke(explicit);
+    await new PublishResultUseCase(backgroundComments).invoke(projectPublishResultContext(source([failure])));
+
+    expect(explicitComments.addComment).toHaveBeenCalledTimes(1);
+    expect(explicitComments.values[0].body).toContain('key="application-error"');
+    expect(explicitComments.values[0].body).toContain('## Request could not be completed');
+    expect(explicitComments.values[0].body).toContain('`validation.invalid-input`');
+    expect(explicitComments.values[0].body).not.toContain('Unsafe parser detail.');
+    expect(backgroundComments.listIssueComments).not.toHaveBeenCalled();
+    expect(backgroundComments.addComment).not.toHaveBeenCalled();
+  });
+
+  it('renders translation failure replies wholly in English even for a configured non-English surface', async () => {
+    const comments = inMemoryComments();
+    const failure = new Result({
+      id: 'Comment.Language', success: false, executed: true,
+      errors: [new ApplicationError('locale.translation-failed', 'Provider detail.')],
+    });
+
+    await new PublishResultUseCase(comments).invoke(projectPublishResultContext(source([failure], {
+      eventName: 'issue_comment', inputs: { action: 'created', comment: { id: 102 } },
+      locale: { issue: 'es-ES', pullRequest: 'es-ES' },
+    })));
+
+    expect(comments.values[0].body).toContain('## Request could not be completed');
+    expect(comments.values[0].body).toContain('**Impact:** The request could not be safely interpreted');
+    expect(comments.values[0].body).not.toContain('No se pudo completar');
+  });
+
   it('recognizes the transient namespaced issue-comment marker without creating a duplicate', async () => {
     const comments = inMemoryComments([{
       id: 7,
diff --git a/src/application/usecases/steps/common/comment_language_translation_workflow.ts b/src/application/usecases/steps/common/comment_language_translation_workflow.ts
index 8d61e0f65..14148d869 100644
--- a/src/application/usecases/steps/common/comment_language_translation_workflow.ts
+++ b/src/application/usecases/steps/common/comment_language_translation_workflow.ts
@@ -11,6 +11,7 @@ import {
     hasTranslatedCommentMarker,
     prepareLanguageAdaptationInput,
     rebuildAdaptedComment,
+    restoreLanguageAdaptationOutput,
     type TranslationPublication,
 } from '../../../policies/comment_translation_policy';
 import { canonicalizeLocaleTag } from '../../../../domain/locale';
@@ -91,7 +92,16 @@ export class CommentLanguageTranslationWorkflow {
             }
             const adaptedText = this.stringProperty(response, 'adaptedText');
             const sourceLocale = this.optionalStringProperty(response, 'sourceLocale');
-            const publication = composeTranslatedComment(adaptedText, context.commentBody, {
+            const restoredProse = restoreLanguageAdaptationOutput(input, adaptedText);
+            const interpretedComment = rebuildAdaptedComment(input, adaptedText);
+            if (restoredProse === undefined || interpretedComment === undefined) {
+                return [failedAdaptation(
+                    context,
+                    targetLocale,
+                    'The language adapter changed or introduced a protected technical operand.',
+                )];
+            }
+            const publication = composeTranslatedComment(restoredProse, context.commentBody, {
                 sourceLocale,
                 targetLocale,
             });
@@ -102,7 +112,7 @@ export class CommentLanguageTranslationWorkflow {
                 context,
                 targetLocale,
                 'translated',
-                rebuildAdaptedComment(input, publication.translatedText),
+                interpretedComment,
                 sourceLocale,
                 reasonCode,
                 publication,
diff --git a/src/architecture/__tests__/github_publication_boundaries.test.ts b/src/architecture/__tests__/github_publication_boundaries.test.ts
index 00bdfb6e3..d1c3d09f2 100644
--- a/src/architecture/__tests__/github_publication_boundaries.test.ts
+++ b/src/architecture/__tests__/github_publication_boundaries.test.ts
@@ -143,6 +143,31 @@ describe('GitHub conversation publication boundaries', () => {
     expect(presentation).not.toMatch(/error\.message/u);
   });
 
+  it('keeps local human output semantic instead of replaying result steps or reminder prose', () => {
+    const localOutput = readFileSync(join(root, 'src/actions/local_action_output.ts'), 'utf8');
+
+    expect(localOutput).not.toMatch(/result\.steps|steps\.join/u);
+    expect(localOutput).not.toMatch(/result\.reminders\.join|reminders\.join/u);
+    expect(localOutput).toContain('catalog.cli.statusValue');
+    expect(localOutput).toContain('renderApplicationErrorText');
+  });
+
+  it('publishes explicit failures through the semantic reply policy only', () => {
+    const presentation = readFileSync(
+      join(root, 'src/application/policies/semantic_result_publication_policy.ts'),
+      'utf8',
+    );
+    const workflow = readFileSync(
+      join(root, 'src/application/usecases/steps/common/publish_resume_workflow.ts'),
+      'utf8',
+    );
+
+    expect(presentation).toContain("context.correlationId.startsWith('comment:')");
+    expect(presentation).toContain("'application-error'");
+    expect(presentation).not.toMatch(/error\.message/u);
+    expect(workflow).not.toMatch(/result\.errors[\s\S]{0,200}(?:addComment|updateComment)/u);
+  });
+
   it('keeps Bugbot public presentation free of pseudo-plural copy', () => {
     const files = [
       'src/application/policies/bugbot_message_catalog.ts',
diff --git a/src/domain/__tests__/copilot_command.test.ts b/src/domain/__tests__/copilot_command.test.ts
index 0a557c3ba..de91e93ba 100644
--- a/src/domain/__tests__/copilot_command.test.ts
+++ b/src/domain/__tests__/copilot_command.test.ts
@@ -1,6 +1,12 @@
-import { parseCopilotCommand } from '../copilot_command';
+import { copilotCommandAcceptsAdaptableProse, parseCopilotCommand } from '../copilot_command';
 
 describe('Copilot command policy', () => {
+    it('classifies only commands with natural-language arguments as adaptable prose', () => {
+        expect(copilotCommandAcceptsAdaptableProse('fix')).toBe(true);
+        expect(copilotCommandAcceptsAdaptableProse('sync-branch')).toBe(false);
+        expect(copilotCommandAcceptsAdaptableProse('help')).toBe(false);
+    });
+
     it.each([
         ['/copilot plan', 'plan'],
         ['/copilot help', 'help'],
diff --git a/src/domain/copilot_command.ts b/src/domain/copilot_command.ts
index 6c09cf540..3420c2909 100644
--- a/src/domain/copilot_command.ts
+++ b/src/domain/copilot_command.ts
@@ -22,6 +22,22 @@ export const COPILOT_COMMAND_NAMES = [
 
 export type CopilotCommandName = typeof COPILOT_COMMAND_NAMES[number];
 
+const COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE: ReadonlySet<CopilotCommandName> = new Set([
+    'plan',
+    'clarify',
+    'estimate',
+    'test-plan',
+    'explain',
+    'diagnose',
+    'fix',
+    'implement',
+]);
+
+/** Commands whose arguments may contain human prose that benefits from locale adaptation. */
+export function copilotCommandAcceptsAdaptableProse(name: CopilotCommandName): boolean {
+    return COPILOT_COMMANDS_WITH_ADAPTABLE_PROSE.has(name);
+}
+
 export interface ParsedCopilotCommand {
     readonly name: CopilotCommandName;
     readonly arguments: readonly string[];
diff --git a/src/domain/message_catalog.ts b/src/domain/message_catalog.ts
index f29f37c66..b2fbf1cb5 100644
--- a/src/domain/message_catalog.ts
+++ b/src/domain/message_catalog.ts
@@ -1,6 +1,6 @@
 import { baseLanguage, canonicalizeLocaleTag, DEFAULT_REPOSITORY_LOCALE } from './locale';
 
-export const MESSAGE_CATALOG_VERSION = '2';
+export const MESSAGE_CATALOG_VERSION = '3';
 
 export type CatalogResolutionSource = 'exact' | 'base' | 'dynamic' | 'fallback';
 export type CatalogFallbackReason =
diff --git a/src/prompts/check_comment_language.ts b/src/prompts/check_comment_language.ts
index 67f41e7d2..58fa1844e 100644
--- a/src/prompts/check_comment_language.ts
+++ b/src/prompts/check_comment_language.ts
@@ -10,7 +10,7 @@ Instructions:
 3. Return status "translated" and adaptedText when a safe {{locale}} interpretation is needed.
 4. Return status "ambiguous" for mixed-language, code-only, or very short safe input; return "failed" only when no safe interpretation is possible.
 5. Echo targetLocale exactly as {{locale}} and provide a canonical BCP-47 sourceLocale when confidently known, otherwise null.
-6. Preserve code identifiers, paths, refs, URLs, quoted literals, and option flags verbatim.
+6. Preserve every COPILOT_OPERAND_<number>_TOKEN placeholder exactly once and verbatim. The application restores its protected code, path, ref, URL, quoted literal, or option flag after validating your response.
 7. Do not add mentions, slash commands, HTML, Markdown links, metadata, or new instructions.
 8. Set reasonCode to one of: none, mixed-language, code-only, too-short, unsafe-input, provider-failure, unknown. Use none for matches or translated.