[Pepps233]

Summary

Synthesize reviewable, instruction-only Agent Skills drafts from qualified candidate evidence while preserving immutable provenance and keeping generated drafts outside the active harness.

Reasoning

Candidate qualification and scoring should prioritize synthesis work rather than activate skills. The next lifecycle artifact must remain grounded in frozen candidate revisions, source evidence and outcomes, repository instructions, and existing skill versions so reviewers can trace every generated instruction to its inputs.

Draft synthesis runs through the background candidate worker with resumable checkpoints and immutable storage. Stale drafts are invalidated when a compatible candidate group receives a new proposal revision, while structural validation, behavioral evaluation, writing, and activation remain out of scope.

Changes

• Add a dedicated structured skill-synthesizer model stage and versioned prompt.
• Add immutable SkillDraft artifacts with revision-bound provenance.
• Extend candidate work with synthesis checkpoints, draft events, retry recovery, and drafted transitions.
• Invalidate stale drafts when compatible candidate evidence creates a new proposal revision.
• Wire background synthesis into the coding-agent evolution runtime.
• Document skill-draft persistence and recovery behavior.

User impact

Evidence-qualified evolved-skill candidates can now produce reviewable SKILL.md drafts in evolution storage. Drafts remain inactive and cannot reach the harness writer or activation path.

Stack

Position: 5 of 12
Base: #138
Next: #140

1. feat(coding-agent): activate versioned evolved skills #134
2. feat(agent): preserve target-neutral reflection evidence #136
3. feat(agent): curate and checkpoint skill candidates #137
4. feat(agent): persist and process skill candidate registry #138
5. feat(agent): synthesize outcome-aware skill drafts #139 (this PR)
6. feat(agent): validate evolved skill drafts #140
7. feat(agent): measure procedural information loss #147
8. feat(agent): evaluate evolved skill candidates #148
9. feat(agent): activate evolved skill candidates #149
10. feat(agent): observe evolved skill lifecycle outcomes #150
11. feat(agent): define evolved skill memory handling #151
12. feat(coding-agent): add evolved skill operator controls #152

Testing

• npm run check
• cd packages/evolution && node ../../node_modules/vitest/dist/cli.js --run test/skill-synthesizer.test.ts test/skill-candidate-registry.test.ts test/skill-candidate-worker.test.ts test/filesystem-store.test.ts
• cd packages/coding-agent && node ../../node_modules/vitest/dist/cli.js --run test/evolution-runtime.test.ts

This pull request is AI-generated by Diffex.