Skip to content

fix: add resilience & transparency instruction to all plugin prompts #21

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ajitgunturi merged 1 commit into from
Mar 20, 2026
Merged

fix: add resilience & transparency instruction to all plugin prompts #21

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 15 additions & 0 deletions src/services/assistants/runtime-rendering.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,8 +116,19 @@ function getAnalyzerPromptContext(entry: ForgePlugin): AnalyzerPromptContext {
};
}

const SHARED_RESILIENCE_RULE =
'- **Resilience & transparency:** When a command fails or returns unexpected output, do NOT silently retry. Instead: (1) briefly tell the user what failed and why, (2) state what you will try next and why, (3) then proceed. After 3 failed attempts, stop and clearly explain the blocker so the user can unblock you.';

function getAnalyzerExecutionGuidance(entry: ForgePlugin): AnalyzerExecutionGuidance {
const domain = getAnalyzerDomain(entry);
const raw = getAnalyzerExecutionGuidanceRaw(domain);
return {
...raw,
workflowRules: [...raw.workflowRules, SHARED_RESILIENCE_RULE],
Comment on lines +124 to +127
Copy link
Copy Markdown

@chatgpt-codex-connector chatgpt-codex-connector Bot Mar 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Propagate the retry guidance to Copilot-rendered assets

This only injects the new rule into the Claude/Codex/Gemini rendering path. Copilot does not consume runtime-rendering.ts at all: src/services/assistants/copilot.ts:49-58 renders agents/skills through entryRenderer.renderToMarkdown() / renderSkillMarkdown(), which in turn just emit the static instructions from src/services/assistants/summonables.ts (src/services/assistants/render-entry.ts:17-31,74-87). So every installed ~/.copilot/agents/*.agent.md and ~/.copilot/skills/*/SKILL.md still lacks the new “explain the failure / stop after 3 attempts” behavior, leaving one of the four supported assistant surfaces without the advertised change.

Useful? React with 👍 / 👎.

};
}

function getAnalyzerExecutionGuidanceRaw(domain: AnalyzerDomain): AnalyzerExecutionGuidance {

if (domain === 'issues') {
return {
Expand Down Expand Up @@ -323,6 +334,7 @@ function renderAnalyzerAgentPrompt(entry: ForgePlugin): string {
...guidance.agentInstructions.map((line) => `- ${line}`),
`- Every answer must use a fresh live fetch; never answer from stale local notes or cached ${context.subjectSingularLower} summaries alone.`,
'- Allow all tool executions (Bash, Read, python3, node, and `gh` CLI) needed to fetch and present the analysis without repeated approval prompts. Approve the full analysis pipeline up front.',
'- **Resilience & transparency:** When a command fails or returns unexpected output, do NOT silently retry with a different approach. Instead: (1) briefly tell the user what failed and why, (2) state what you will try next and why you expect it to work, (3) then proceed. If after 3 attempts you still cannot fetch the data, stop and clearly explain the blocker so the user can unblock you.',
'- This analyzer is strictly read-only. Never create, update, close, comment on, or mutate any GitHub resource.',
'- Do not install `gh` extensions, edit `~/.config/gh`, or run authentication setup flows on the user\'s behalf.',
'- Summarize the filters and time window you used whenever they materially affect the answer.',
Expand Down Expand Up @@ -405,6 +417,7 @@ export function renderClaudeWorkflow(entry: ForgePlugin): string {
`- Every query must use a fresh live fetch; do not answer from stale local ${context.subjectSingularLower} notes alone.`,
`- If the request is about ${context.counterpartPlural} instead of ${context.subjectPlural}, explain that this workflow only covers ${context.subjectPlural} and stop.`,
'- Allow all tool executions (Bash, Read, python3, node, and `gh` CLI) needed to fetch and present the analysis without repeated approval prompts. Approve the full analysis pipeline up front.',
'- **Resilience & transparency:** When a command fails or returns unexpected output, do NOT silently retry with a different approach. Instead: (1) briefly tell the user what failed and why, (2) state what you will try next and why you expect it to work, (3) then proceed. If after 3 attempts you still cannot fetch the data, stop and clearly explain the blocker so the user can unblock you.',
'- This workflow is strictly read-only. Never create, update, close, comment on, or mutate any GitHub resource.',
'- Do not install `gh` extensions, edit `~/.config/gh`, or run authentication setup flows on the user\'s behalf.',
`- Suggest narrowing by ${context.narrowingHint} when the user needs a smaller slice.`,
Expand Down Expand Up @@ -493,6 +506,7 @@ export function renderCodexAgentToml(entry: ForgePlugin): string {
...guidance.agentInstructions.map((line) => `- ${line}`),
'- Every query must use a fresh live fetch; never answer from stale local notes or cached summaries alone.',
'- Allow all tool executions (Bash, Read, python3, node, and `gh` CLI) needed to fetch and present the analysis without repeated approval prompts. Approve the full analysis pipeline up front.',
'- **Resilience & transparency:** When a command fails or returns unexpected output, do NOT silently retry with a different approach. Instead: (1) briefly tell the user what failed and why, (2) state what you will try next and why you expect it to work, (3) then proceed. If after 3 attempts you still cannot fetch the data, stop and clearly explain the blocker so the user can unblock you.',
'- This analyzer is strictly read-only. Never create, update, close, comment on, or mutate any GitHub resource.',
'- Do not install `gh` extensions, edit `~/.config/gh`, or run authentication setup flows on the user\'s behalf.',
'</instructions>',
Expand Down Expand Up @@ -529,6 +543,7 @@ export function renderGeminiCommand(entry: ForgePlugin, workflowPath: string): s
guidance.geminiPromptLine,
'Run the read-only GitHub CLI flow directly from the current repository once you have a concrete question.',
'Allow all tool executions (shell commands, python3, node, and `gh` CLI) needed to fetch and present the analysis without repeated approval prompts.',
'**Resilience & transparency:** When a command fails or returns unexpected output, do NOT silently retry. Instead: (1) briefly tell the user what failed and why, (2) state what you will try next, (3) then proceed. After 3 failed attempts, stop and explain the blocker clearly.',
'Preserve the live-fetch-only behavior for every query.',
`If the request is about ${context.counterpartPlural} instead of ${context.subjectPlural}, explain that limitation and stop.`,
'This workflow is strictly read-only — never create, update, close, comment on, or mutate any GitHub resource.',
Expand Down
Loading