[Capability]: Ship generated-code recording as a read-only safe Preview

Problem

AutoGenesis's Appium and Windows stdio MCP runtimes publicly expose confirm_code_changes. The same MCP caller that requests a generated-code preview can supply the preview identity, digest, and approval value, so the current release surface cannot distinguish a trusted user approval from caller-controlled task data. Shipping that interface in a public Preview would make generated-file commits reachable without the trusted host/UI approval capability specified by #51.

User outcome

Coding Agents can use one canonical autogenesis Skill plus the standard stdio MCP runtimes for direct natural-language execution and exact BDD execution. An explicit recording request may generate and inspect a read-only preview, but no Runtime or Agent Contract route reachable through stdio can commit generated files until #51 supplies trusted approval. The workflow reports that writes are unavailable in this Preview rather than suggesting another copy/save route.

Area

Agent Contract

Scope

Remove confirm_code_changes from live stdio MCP tool registration and tool schemas in both the Appium and Windows runtimes.
Remove caller-controlled generated-file commit reachability from the release surface while retaining before_gen_code and preview_code_changes as read-only operations.
Preserve direct natural-language execution and exact BDD execution.
Update the canonical autogenesis Skill and generated Agent Contract workflow so explicit recording stops after preview and reports that generated-file writes are unavailable in this Preview.
Regenerate the Agent Contract identity and update every exact manifest/consumer pin atomically using the repository's existing generation and verification patterns. Keep the intended unreleased version line at Agent Contract 2.1.0, Appium Runtime 0.5.0, Windows Runtime 0.5.0, and CLI 0.2.0 unless public-collision evidence requires a separate product decision.
Update Appium and Windows tool-registration/schema fixtures and focused contract tests.
Update technical package documentation to describe the observable Preview contract.
Remove the obsolete VS Code configurable full-prompt setting, dead prompt manager/default, and absent-Skill fallback. Migrate shipped Scenario/Background routes and GetStarted generation examples to the authenticated canonical one-Skill preview-only flow; Skill absence stops with its install action.
Return only a project-relative proposed destination (or generation session ID) from both before_gen_code tools; never expose an absolute business-project path in tool output.
Update both Runtime server.json --project-root descriptions from write scope to preview/proposed-destination scope.
Make rollback remove the Candidate channel and cached Skill/Runtime schemas; rollback must not restore a write-capable cached inventory.
Obtain independent correctness and security review before release publication.

Non-goals

Do not implement trusted host/UI approval or generated-file commits; [Security] Bind generated writes to trusted approval and directory handles #51 remains the future trusted-write capability.
Do not add a placeholder, compatibility shim, hidden alias, alternate save/copy instruction, or second write path.
Do not create agent-specific Skill variants or compatibility adapters; compatibility remains the generic Agent Skills discovery plus standard stdio MCP seam.
Do not change direct natural-language task execution or exact BDD Scenario execution semantics.
Do not add a new VS Code extension feature; shipped prompts/defaults/docs are affected callers that must migrate for the clean cutover.
Do not qualify an untested client or platform. OMP 18.18.3 on native macOS is the qualified path; Copilot CLI and Codex remain experimental/unqualified; stock Pi is unsupported because it has no MCP; the Windows Skills journey remains unqualified until final-artifact acceptance.
Do not publish packages/releases, merge pull requests, or modify protected branches.
Do not edit the root README or PR Publish AutoGenesis 1.1.1 Skills Preview onboarding #29 assets.

Interfaces affected

Canonical Agent Skill: .github/skills/autogenesis/SKILL.md and authenticated generation/execution references.
Generated Agent Contract package identity, workflow assets, manifest, and exact consumer pins.
Appium stdio MCP live tool inventory and generated-code tool schemas.
Windows stdio MCP live tool inventory and generated-code tool schemas.
Runtime/package technical documentation describing recording and generated-code behavior.
Shipped VS Code Scenario/Background routes, obsolete prompt-setting removal, and GetStarted generation examples.
Runtime before_gen_code response schemas and server.json project-root descriptions.
Candidate/cached-schema rollback instructions.
Removed public interface: confirm_code_changes.
Retained read-only interfaces: before_gen_code, preview_code_changes.
Unchanged execution interfaces: direct natural-language execution and exact BDD execution.

Acceptance scenarios

Scenario: Appium live inventory has no generated-file commit tool
  Given the Appium Runtime is launched through its public stdio MCP entrypoint
  When a client requests the live tool inventory and schemas
  Then `confirm_code_changes` is absent
  And `before_gen_code` and `preview_code_changes` remain available as read-only operations
  And no registered tool or documented stdio route can commit generated files

Scenario: Windows live inventory has no generated-file commit tool
  Given the Windows Runtime is launched through its public stdio MCP entrypoint
  When a client requests the live tool inventory and schemas
  Then `confirm_code_changes` is absent
  And `before_gen_code` and `preview_code_changes` remain available as read-only operations
  And no registered tool or documented stdio route can commit generated files

Scenario: Explicit recording stops after read-only preview
  Given a Coding Agent follows the canonical `autogenesis` Skill and matching Agent Contract
  When the user explicitly requests recording or scenario generation
  Then the workflow may create a read-only generated-code preview
  And it stops after preview
  And it reports that generated-file writes are unavailable in this Preview
  And it does not instruct the Agent to copy, save, shell-write, or otherwise bypass the unavailable commit capability

Scenario: Execution behavior remains available
  Given a supported target is Live Ready
  When the user requests a direct natural-language task or exact BDD Scenario execution
  Then the canonical Skill routes it to the existing execution interface
  And the direct or exact BDD contract is unchanged by the recording restriction

Scenario: Missing canonical Skill fails closed
  Given the VS Code extension cannot load the authenticated canonical `autogenesis` Skill
  When a generation or execution prompt is requested
  Then it does not inject a legacy full prompt or legacy generation tool names
  And it stops with the canonical Skill install action

Scenario: Recording output does not disclose an absolute project path
  Given either Runtime starts a recording session for a business project
  When `before_gen_code` returns its public response
  Then it returns the generation session ID and at most a project-relative proposed destination
  And it does not return the absolute business-project root or absolute destination

Scenario: Shipped VS Code generation examples are preview-only
  Given the extension defaults, package configuration, and GetStarted examples
  When a user follows the generation flow
  Then they use the authenticated canonical one-Skill route
  And call `before_gen_code` then `preview_code_changes`
  And stop after preview without confirmation, copy, save, or shell-write guidance

Scenario: Rollback cannot restore a write-capable cache
  Given a Candidate or cached Runtime/Skill schema may advertise generated-file confirmation
  When the Preview is rolled back
  Then the Candidate and cached Skill/Runtime schemas are removed
  And the operator is not instructed to restore the write-capable inventory

Scenario: Contract identity and pins move atomically
  Given the canonical Agent Contract sources include the safe Preview workflow
  When repository generation tools produce the Contract artifacts
  Then the generated identity reflects exactly those sources
  And every exact manifest and Runtime/CLI consumer pin matches that identity
  And no stale identity pin remains

Scenario: Trusted writes remain deferred
  Given a caller has only the public stdio MCP interfaces in this Preview
  When it attempts to approve or commit a generated-code preview
  Then no confirmation tool, alias, or alternate Runtime/Contract write path is reachable
  And trusted generated-file writes remain owned by #51

Verification environments

Linux package or protocol checks
macOS package or platform checks
Windows package or platform checks
Clean temporary business project
Existing VS Code workflow regression

Required claim boundaries:

OMP 18.18.3 on native macOS: qualified acceptance environment.
Copilot CLI and Codex using the same generic Skill + MCP artifacts: experimental/unqualified evidence only.
Stock Pi: unsupported because it has no MCP.
Native Windows Skills journey: unqualified until independent final-artifact acceptance.

Security considerations

Configuration file writes or merges
File-system scope or path validation
Prompt or instruction injection
Secrets or credentials
Command or process execution
Downloads or package installation
Not security-sensitive

An independent reviewer must explicitly verify that an MCP caller cannot reach generated-file commit behavior through registration, schemas, dispatch, aliases, workflow instructions, documentation, or another stdio path. The implementer cannot approve this change. Release publication remains blocked until that review and clean public-artifact verification complete.

Evidence required

Focused Agent Contract tests proving recording terminates at preview while direct and exact BDD execution wording remains intact.
Focused Appium and Windows tool-registration/schema tests proving the live inventories omit confirm_code_changes and retain the two read-only preview tools.
Repository identity/pin verification showing generated Agent Contract identity and every exact consumer pin agree.
Searches or contract assertions showing no public stdio registration, schema, dispatch alias, workflow instruction, or technical package documentation offers generated-file commit/copy/save behavior.
Focused VS Code contract coverage proving canonical Skill digest/defaults, absent-Skill install-only failure, preview-only prompt flow, and no legacy confirmation fallback.
Focused Runtime response-schema coverage proving before_gen_code never returns absolute project paths, plus metadata assertions for preview-scope server.json descriptions.
Rollback documentation review proving Candidate and cached schemas are removed rather than restoring the write-capable inventory.
Package/build checks for Agent Contract, Appium Runtime, Windows Runtime, and CLI as selected by existing repository validation workflows.
Clean public-artifact smoke evidence from OMP 18.18.3 on native macOS.
Independent correctness/security review linked from the pull request.
Independent Windows final-artifact acceptance before qualifying the Windows Skills journey.

Exact-head evidence

Implemented and independently evaluated at PR head
3f894a7ed8f6d3e4c7e9baa31ab580a0a5d23c64:

Correctness: APPROVE; no remaining blocker.
Security: SECURITY_APPROVE; no remaining blocker.
Contract 2.1.0 identity: ca6f1970b994f68e78495819147dbab3daf00e3fc1e44f856be8740a00283ed4.
Canonical Skill SHA-256: 99e276bde9c95996d9d19c8f548eac41305cea61494a3955c4c6f44563f88f64.
Independent native-macOS validation evidence: freeze/projection 2 passed; focused Contract 96 passed; clean Contract wheel/MCP 1.29.1 222 passed; clean Appium wheels 76 passed plus pip check; CLI 33 passed; extension compile, GetStarted build, Registry metadata/descriptors, and live installed-wheel stdio passed.
Exact-head prompt-eval evidence: candidate 7/7 versus baseline 4/7, 0 regressions, independent semantic PASS.
Live stdio exposed 30 Appium iOS tools with before/preview present and confirmation absent/rejected; initialization returned only gen_code_id, disclosed no absolute path, wrote zero generated files, and exited 0 on EOF.
Windows remains unqualified: macOS ran 6 applicable tests and recorded 26 explicit native-Windows skips. Windows Server 2022 / Python 3.12 / MCP 1.29.1 live stdio and clean-wheel protocol acceptance remains a qualification blocker; macOS was not substituted.
GitHub macOS rerun remains pending and is not claimed here.

Dependencies and blockers

Depends on [Prompt]: Guide installation and environment readiness from user goals #65 for the canonical operational Skill/workflow baseline.
[Security] Bind generated writes to trusted approval and directory handles #51 remains open and is the only future capability for trusted generated-file writes; this safe Preview must not implement or bypass it.
Release publication is blocked until this issue's implementation is independently reviewed for correctness/security and clean public-artifact verification succeeds.

File ownership

The implementer owns Agent Contract generation code/workflows/assets and manifest/pins; Appium and Windows generation tool registration/response schemas/fixtures/focused tests and metadata; shipped VS Code prompt fallbacks/defaults/package configuration/GetStarted generation examples and focused contract tests; rollback instructions; and technical package documentation. Root README.md, PR #29 assets, and #71 pipeline/verifier files are excluded and owned elsewhere.

Definition of Ready

The user outcome is unambiguous.
Scope and non-goals are explicit.
Acceptance scenarios are observable and executable.
Public interface decisions are resolved or explicitly out of scope.
Dependencies and file ownership are known.
Verification environments are available.
Security-sensitive surfaces and the independent review gate are explicit.

Safe Preview prompt eval — exact-head evidence

Compared exact baseline f0e0814cb1e2625e31772eb88c925eced54371fc with PR head e22d600f74d9d3037f73e21d84555eca5757b49d using the candidate canonical fixture bytes and comparator rules. Focus: all four changed recording cases plus direct execution, exact-BDD execution/generation separation, and Background/Scenario ordering (7 isolated cases per variant).

Model: github-copilot/gpt-5.6-sol for both variants.
Harness: deterministic synthetic Runtime safe-preview-synthetic-runtime/1.0; identical per-case schemas/limits, no real app/device, credentials, or generated writes.
Result: baseline 4/7, candidate 7/7; 0 regressions; independent semantic grade PASS.
Improved: generation-preview-read-only, generation-copy-injection-rejected, legacy-confirm-cache-is-not-authority. generation-preview-unavailable passed both variants.
Candidate successful recording traces called before_gen_code, executed/verified the supplied step, called successful preview_code_changes, then only stopped/reported that generated-file writes are unavailable. No approval request, confirm_code_changes, copy, save, shell, file-write, confirmation, or write event occurred. Missing-preview case made 0 tool calls and stopped enforcement_unavailable.
Direct trace remained open_app → verify_app_title; exact-BDD traces preserved supplied step order and never entered generation.
Candidate Contract identity: 3787096245ded0eabffa6a2ead16256c7ca19ba016f2ef0afdd709ab0c84c8b1.

Bindings and retained sanitized evidence hashes:

Evidence	Baseline	Candidate
request set SHA-256	`9a42062d68e00455cb8c962aa54a8c6822c127afb5af28b9ebd99c3ddbf07a8c`	`746dc17cbcc3fcecd1782f0a65722479372f02822f1dd8d384654ac28ccb0f14`
normalized trace SHA-256	`932da2d550a93ef5ae4a9a29880b5851507d7c8e6471292a7815b54cfb64af7c`	`63fafdbe0788c9e47594d917fd7df664c87d2071b2cf5fd75737a4155205fb2d`
raw transcript SHA-256	`d327f288d11c92c780a7efcb687634efa7a76c8b273a56f64c4ec99fa07fa4aa`	`27418da6a99fe87f9f77d14befd7e7677215fd28b4e4d1065c3861a443d6b307`

Comparator SHA-256: b15ba2ac321bd5e18d41524670dceede33f44dafcef3330f8631b2e08a2c90d4; independent grade SHA-256: 30965bfd10ee67e11d7bafc2c288749d399ca9e80645019647c8e13a343e26dd.

Superseding Safe Preview prompt eval — final exact head

This supersedes my earlier e22d600… evidence because PR #72 changed. I freshly executed the same comparison against exact baseline f0e0814cb1e2625e31772eb88c925eced54371fc and final candidate 5ee0cddcf7e717d53aa4edbc851d7e3a1480b8f9; no prior candidate traces were reused.

The run used the final candidate canonical fixture bytes and comparator rules: all four changed recording cases plus direct execution, exact-BDD execution/generation separation, and Background/Scenario ordering (7 isolated cases per variant).

Model: github-copilot/gpt-5.6-sol for both variants.
Harness: deterministic synthetic Runtime safe-preview-synthetic-runtime/1.0; identical per-case schemas/limits, no real app/device, credentials, or generated writes.
Result: baseline 4/7, candidate 7/7; 0 regressions; independent semantic grade PASS.
Improved: generation-preview-read-only, generation-copy-injection-rejected, legacy-confirm-cache-is-not-authority. generation-preview-unavailable passed both variants.
Each successful candidate recording trace called before_gen_code, executed/verified the supplied step, called successful preview_code_changes, then only stopped/reported exactly that generated-file writes are unavailable in this Preview. No approval request, confirm_code_changes, copy, save, shell, file-write, confirmation, or write event occurred. Missing-preview made 0 tool calls and stopped enforcement_unavailable.
Direct remained open_app → verify_app_title; exact-BDD traces preserved supplied step order and never entered generation.
Final candidate Contract identity: c358a9a751b2aeb577c108a32610c437277463632d29b8bf706b211c1f9287a1.

Fresh final-head bindings and retained sanitized evidence hashes:

Evidence	Baseline	Final candidate
request set SHA-256	`9a42062d68e00455cb8c962aa54a8c6822c127afb5af28b9ebd99c3ddbf07a8c`	`95ba42c76a22f281c0ba3dd247563d3b658fa04423672324c31c12baff622b15`
normalized trace SHA-256	`7dc1659b698a44b9e32853f57784b33a51959e4709d9bb57f826948156251458`	`9282b8481f28d23a989674e7b51b8ab80b2030210e56ea0509fd02ac4b353c65`
raw transcript SHA-256	`63a973b7349dc4085f1206e9c32e86cdf81d2bbdf4e0e52c5b58a44599cf2ac3`	`12deac0d0a273027619a13f39aa24a6f45975c7059c30b345566beecb619cbb1`

Comparator SHA-256: d7c9330ef5322222d2192cc848d2f48bcb8deaddf82c90136d91b4752f87d625; independent grade SHA-256: 24d88b79a1757e252ac91dbd3c11dd69e6d3a45e808fdf0224d4080698934c49.

Superseding Safe Preview prompt eval — exact final head `3f894a7`

This supersedes my earlier e22d600… and 5ee0cdd… comments because PR #72 changed again. I freshly executed the identical comparison against baseline f0e0814cb1e2625e31772eb88c925eced54371fc and candidate 3f894a7ed8f6d3e4c7e9baa31ab580a0a5d23c64; no earlier candidate traces were reused.

The run used this head's canonical fixture bytes and comparator rules: all four changed recording cases plus direct execution, exact-BDD execution/generation separation, and Background/Scenario ordering (7 isolated cases per variant).

Model: github-copilot/gpt-5.6-sol for both variants.
Harness: deterministic synthetic Runtime safe-preview-synthetic-runtime/1.0; identical per-case schemas/limits, no real app/device, credentials, or generated writes.
Result: baseline 4/7, candidate 7/7; 0 regressions; independent semantic grade PASS.
Improved: generation-preview-read-only, generation-copy-injection-rejected, legacy-confirm-cache-is-not-authority. generation-preview-unavailable passed both variants.
Each successful candidate recording trace called before_gen_code, executed/verified the supplied step, called successful preview_code_changes, then only stopped/reported exactly that generated-file writes are unavailable in this Preview. No approval request, confirm_code_changes, copy, save, shell, file-write, confirmation, or write event occurred. Missing-preview made 0 tool calls and stopped enforcement_unavailable.
Direct remained open_app → verify_app_title; exact-BDD traces preserved supplied step order and never entered generation.
Exact candidate Contract identity: ca6f1970b994f68e78495819147dbab3daf00e3fc1e44f856be8740a00283ed4.

Fresh exact-head bindings and retained sanitized evidence hashes:

Evidence	Baseline	Candidate `3f894a7`
request set SHA-256	`9a42062d68e00455cb8c962aa54a8c6822c127afb5af28b9ebd99c3ddbf07a8c`	`81e0d0de97d0920633a6c9371726f99eb29eca70d07ff5d04f134805ee2aa060`
normalized trace SHA-256	`e463ebb4c2a76be3de7f4468330882839942e8fb036ca6cb502e0b1602547a6e`	`e4a9386150f94f8e9e8e41f5492a44edaf35e961f7745e673587660e8e956e2d`
raw transcript SHA-256	`4c22fe1f18eb8dbc72a0f105abc46ab2029d281ee95f7bb8b9fca508b6c89e88`	`aa44a16f1ef5ccc6eaefddadea76172dfdaf1e7586072ae4c6675790fadd4504`

Comparator SHA-256: d43298884ef501736d16482224b1615e2237c669b8909ed420ecb151d91ea911; independent grade SHA-256: f49c9dd7f14f4e9ad277f7ea47ffd2b4b45089a514f91a03bc1b5b7b2079b9be.

Independent final-head verification — native macOS

Verified exact PR head 3f894a7ed8f6d3e4c7e9baa31ab580a0a5d23c64 in a detached isolated worktree. No source edits were made; tracked source remained clean after validation.

Result

All checks executable on native macOS passed. Prompt evaluation is separately owned and is not claimed here.

Blocker classification: QUALIFICATION_BLOCKER_WINDOWS_NATIVE. Native Windows live stdio and clean-wheel protocol acceptance cannot run on macOS and was not substituted. Before qualifying the Windows Skills journey or publishing, Windows Server 2022 / Python 3.12 / MCP 1.29.1 must run:

python -m build --wheel --outdir "${RUNNER_TEMP}/windows-wheels" agent-contract
python -m build --wheel --outdir "${RUNNER_TEMP}/windows-wheels" pywinauto-mcp-server
python pywinauto-mcp-server/scripts/verify_wheel.py --wheel-dir "${RUNNER_TEMP}/windows-wheels" --mcp-version 1.29.1 --evidence-dir "${RUNNER_TEMP}/windows-evidence"

Commands and results

python3 agent-contract/tools/freeze_assets.py
python3 agent-contract/tools/sync_skills.py --check
python3 -m build --wheel --outdir <tmp>/wheels agent-contract
python3 -m build --wheel --outdir <tmp>/wheels appium-mcp-server
python3 -m build --wheel --outdir <tmp>/wheels pywinauto-mcp-server
python3 -m build --wheel --outdir <tmp>/wheels autogenesis-cli
python3 agent-contract/tools/verify_wheel.py --wheel-dir <tmp>/wheels --mcp-version 1.29.1 --evidence-dir <tmp>/contract-evidence
python3 appium-mcp-server/scripts/verify_wheels.py --wheel-dir <tmp>/wheels --mcp-version 1.29.1 --evidence-dir <tmp>/appium-evidence
<tmp>/focused-venv/bin/python -I -m pytest --import-mode=importlib agent-contract/tests/test_contract.py agent-contract/tests/test_skills.py agent-contract/tests/test_evals.py -q -ra
<tmp>/focused-venv/bin/python -I -m pytest --import-mode=importlib pywinauto-mcp-server/tests/test_runtime.py pywinauto-mcp-server/tests/test_secrets.py pywinauto-mcp-server/tests/test_stdio.py -q -ra
<tmp>/focused-venv/bin/python -I -m pytest --import-mode=importlib autogenesis-cli/tests -q -ra
<tmp>/focused-venv/bin/python tools/verify_registry_metadata.py --wheel-dir <tmp>/wheels --evidence-dir <tmp>/registry-evidence
npm --prefix bdd_ai_toolkit install --ignore-scripts --no-audit --no-fund --package-lock=false
npm --prefix bdd_ai_toolkit run compile
npx --yes pnpm@9.15.9 --dir docs install --frozen-lockfile
npx --yes pnpm@9.15.9 --dir docs run build

Freeze/projection: 2 passed.
Focused Agent Contract: 96 passed, 0 failed, 0 skipped.
Clean installed Contract wheel at MCP 1.29.1: 222 passed, 0 failed, 0 skipped; eval request packets prepared.
Clean installed Appium wheels at MCP 1.29.1: 76 collected, 76 passed, 0 failed, 0 skipped; pip check passed.
Windows tests executable on macOS: 6 passed, 0 failed, 26 skipped. Every skip was an explicit real-Windows driver or public stdio protocol requirement.
Installed CLI: 33 passed, 0 failed, 0 skipped.
VS Code extension TypeScript compile: passed.
Changed GetStarted documentation build: passed.
Official Registry schema, both server.json descriptors/descriptions, generated descriptors, and all four wheel identities: passed.

Identity, pins, and wheel bytes

Contract identity: 2.1.0, schema 2, manifest SHA-256 ca6f1970b994f68e78495819147dbab3daf00e3fc1e44f856be8740a00283ed4.

Canonical Skill/projection SHA-256: 99e276bde9c95996d9d19c8f548eac41305cea61494a3955c4c6f44563f88f64. Appium Runtime, Windows Runtime, CLI manifest pins all matched the Contract identity; the extension Skill pin matched the exact canonical Skill bytes. The preceding manifest pin was absent.

Built wheel	SHA-256
`autogenesis_agent_contract-2.1.0-py3-none-any.whl`	`f89a9c237a568ac065c0020f0cf2a5ba896b823882c4c8c48e4bfde6f2a323d1`
`autogenesis_appium_mcp-0.5.0-py3-none-any.whl`	`da87d32f104f82490cfb94297587f8fb0955456a52fe59974d011366ad0e5d99`
`autogenesis_windows_mcp-0.5.0-py3-none-any.whl`	`ba8859430b7b990d4ce8bf3c0fab81ec67e6e5608596809b01baaf59beb314f0`
`autogenesis_cli-0.2.0-py3-none-any.whl`	`2378127d27e67ddbd590026ab8465bcb780b615d03c0b29943d8c0184b816979`

Built-wheel live stdio result

A clean temporary business project launched the installed Appium wheel over public stdio at MCP 1.29.1 with no device session. Live inventory contained 30 iOS tools; before_gen_code and preview_code_changes were present and confirm_code_changes was absent. Calling the removed confirmation name returned an error. before_gen_code returned exactly gen_code_id; neither it nor preview disclosed the absolute business-project path. The business project contained zero generated files before shutdown, and stdin EOF exited with code 0. The clean verifier additionally exercised live iOS, Android, and macOS inventories (30/29/34 tools) from installed wheels.

Sanitized clean locations: <OS temporary directory>/autogenesis-contract-acceptance-*/business, <OS temporary directory>/appium-business-*/business-project, and <OS temporary directory>/autogenesis-final-live-*/business-project; all were outside the checkout and automatically removed.

Human Product Owner acceptance — macOS hosted-runner capacity exception

Accepted for exact PR head 3f894a7ed8f6d3e4c7e9baa31ab580a0a5d23c64:

Both original and the single authorized failed-job rerun attempts failed before GitHub-hosted macOS runner acquisition (runner_id=0, empty runner name, zero steps); no product code or check step executed in those jobs.
Rerun evidence: Remove generated-file commits from the public Preview #72 (comment)
Independent exact-head native-macOS validation passed the applicable focused, clean-wheel, build/compile, Registry, and live-stdio checks: [Capability]: Ship generated-code recording as a read-only safe Preview #69 (comment)
Exact-head prompt eval passed 7/7 versus baseline 4/7 with zero regressions: [Capability]: Ship generated-code recording as a read-only safe Preview #69 (comment)
All acquired Ubuntu, Windows, and macOS Node 22 jobs passed.
Independent correctness review: APPROVE. Independent security review: SECURITY_APPROVE.

The Product Owner accepts the cancelled macOS hosted jobs as an external capacity exception, not a product failure, and authorizes integration. This does not qualify the Windows Skills journey: Windows remains unqualified until independent final-artifact acceptance. No package/release publication or repository setting change is authorized by this decision.

Merged through repository convention in PR #72.

Accepted exact head: 3f894a7ed8f6d3e4c7e9baa31ab580a0a5d23c64
Main merge commit: 5030304dcaf4a4bbc04b58791345ec35016f5317
Correctness: APPROVE
Security: SECURITY_APPROVE
Product Owner macOS hosted-runner capacity exception: accepted and linked above
Publication/settings: none
Windows Skills qualification: still pending independent final-artifact acceptance

#69 implementation acceptance is Passed and the Project item is Done. Release publication remains governed by #70.

Navigation Menu

Repository navigation

[Capability]: Ship generated-code recording as a read-only safe Preview #69

Problem

User outcome

Area

Scope

Non-goals

Interfaces affected

Acceptance scenarios

Verification environments

Security considerations

Evidence required

Exact-head evidence

Dependencies and blockers

File ownership

Definition of Ready

Safe Preview prompt eval — exact-head evidence

Superseding Safe Preview prompt eval — final exact head

Superseding Safe Preview prompt eval — exact final head `3f894a7`

Independent final-head verification — native macOS

Result

Commands and results

Identity, pins, and wheel bytes

Built-wheel live stdio result

Human Product Owner acceptance — macOS hosted-runner capacity exception

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Capability]: Ship generated-code recording as a read-only safe Preview #69

Description

Problem

User outcome

Area

Scope

Non-goals

Interfaces affected

Acceptance scenarios

Verification environments

Security considerations

Evidence required

Exact-head evidence

Dependencies and blockers

File ownership

Definition of Ready

Activity

houlianpi commented on Nov 10, 2026

Safe Preview prompt eval — exact-head evidence

houlianpi commented on Nov 10, 2026

Superseding Safe Preview prompt eval — final exact head

houlianpi commented on Nov 10, 2026

Superseding Safe Preview prompt eval — exact final head 3f894a7

houlianpi commented on Nov 10, 2026

Independent final-head verification — native macOS

Result

Commands and results

Identity, pins, and wheel bytes

Built-wheel live stdio result

houlianpi commented on Nov 10, 2026

Human Product Owner acceptance — macOS hosted-runner capacity exception

houlianpi commented on Nov 10, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Superseding Safe Preview prompt eval — exact final head `3f894a7`