stark-ai-de/agent-skills / incubator/skills/skill-maintenance/skillopt-setup
skillopt-setup
FlaggedSet up Microsoft SkillOpt for Agent Skills improvement. Use when installing SkillOpt, preparing per-skill train/val/test eval splits, configuring Codex CLI or provider-backed execution, running SkillOpt against SKILL.md bodies, reviewing best_skill.md, or importing optimized skill changes. Do not use for ordinary skill authoring or manual repo validation.
references/local-artifact-audit.md
View on GitHub# Local Artifact Audit
Use this when prior SkillOpt work under `.agents/` might contain reusable setup improvements.
## Command
```bash
node incubator/skills/skill-maintenance/skillopt-setup/scripts/audit-skillopt-local-artifacts.mjs --skill <skill>
```
Add `--json` when another script or review needs machine-readable output.
## Promotion Rules
Promote only sanitized, reusable material:
- skill-owned helper scripts,
- concise operational references,
- adapter/config templates,
- eval cases,
- curated run summaries under `skill-evals/<skill>/runs/`.
Do not promote these raw local artifacts:
- `.agents/tools/SkillOpt`,
- `.agents/tools/SkillOpt/.git`,
- `.agents/tools/SkillOpt/.venv`,
- `.agents/skills`,
- `.agents/skillopt-work/*/data`,
- `.agents/skillopt-work/*/outputs`,
- `.agents/skillopt-work/_readiness`,
- raw histories, transcripts, auth diagnostics, temporary candidates, or local probe output.
## Required Loop
1. Run the audit before inspecting or moving local `.agents` work.
2. Treat `.agents/skills/<skill>` as an installed copy only; tracked skill source wins.
3. If adapter files differ from tracked templates or the target-specific manifest lacks identity/template freshness metadata, refresh setup before training.
4. Treat `.agents/skillopt-work/adapter-manifest.json` as a legacy compatibility copy when a current target-specific manifest is present; it is not enough by itself for training handoff.
5. Use run summaries only for adoption decisions. A negative held-out test hard-score delta blocks `best_skill.md` adoption.
6. Summary discovery inspects only `.agents/skillopt-work/<skill>/outputs/<run-directory>/summary.json`. It ignores unrelated output files, does not recurse through arbitrary trees, and reports `discovery.status: incomplete` instead of silently trusting a result when a bounded scan limit is reached.
7. Add validator and eval coverage for every promoted behavior.
8. Run repo validation before finalizing tracked changes.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |