skill-meta

Files (49)

• SKILL.md
• SKILL.md · parts/plugin-creator/
• SKILL.md · parts/skill-authoring/
• LICENSE.txt · parts/skill-creator/
• SKILL.md · parts/skill-creator/
• analyzer.md · parts/skill-creator/agents/
• comparator.md · parts/skill-creator/agents/
• grader.md · parts/skill-creator/agents/
• eval_review.html · parts/skill-creator/assets/
• generate_review.py · parts/skill-creator/eval-viewer/
• viewer.html · parts/skill-creator/eval-viewer/
• companion_tooling.md · parts/skill-creator/references/
• description_design_patterns.md · parts/skill-creator/references/
• progressive_disclosure_principles.md · parts/skill-creator/references/
• quality_gates_for_skills.md · parts/skill-creator/references/
• schemas.md · parts/skill-creator/references/
• __init__.py · parts/skill-creator/scripts/
• aggregate_benchmark.py · parts/skill-creator/scripts/
• generate_report.py · parts/skill-creator/scripts/
• improve_description.py · parts/skill-creator/scripts/
• package_skill.py · parts/skill-creator/scripts/
• quick_validate.py · parts/skill-creator/scripts/
• run_eval.py · parts/skill-creator/scripts/
• run_loop.py · parts/skill-creator/scripts/
• skill_description_validator.py · parts/skill-creator/scripts/
• skill_review_checklist_runner.py · parts/skill-creator/scripts/
• skill_structure_validator.py · parts/skill-creator/scripts/
• utils.py · parts/skill-creator/scripts/
• SKILL.md · parts/skill-hub/
• SKILL.md · parts/skill-installer/
• README.md · parts/skill-tester/
• SKILL.md · parts/skill-tester/
• README.md · parts/skill-tester/assets/sample-skill/
• SKILL.md · parts/skill-tester/assets/sample-skill/
• sample_text.txt · parts/skill-tester/assets/sample-skill/assets/
• test_data.csv · parts/skill-tester/assets/sample-skill/assets/
• sample_text_analysis.json · parts/skill-tester/assets/sample-skill/expected_outputs/
• api-reference.md · parts/skill-tester/assets/sample-skill/references/
• text_processor.py · parts/skill-tester/assets/sample-skill/scripts/
• sample_validation_report.json · parts/skill-tester/expected_outputs/
• quality-scoring-rubric.md · parts/skill-tester/references/
• skill-structure-specification.md · parts/skill-tester/references/
• tier-requirements-matrix.md · parts/skill-tester/references/
• audit_skills.py · parts/skill-tester/scripts/
• quality_scorer.py · parts/skill-tester/scripts/
• script_tester.py · parts/skill-tester/scripts/
• security_scorer.py · parts/skill-tester/scripts/
• skill_validator.py · parts/skill-tester/scripts/
• test_security_scorer.py · parts/skill-tester/tests/

SKILL.md (first lines)

# skill-meta — skills-about-skills

> Home branch in Memora: `root/skill/skill-meta`. The layer's law is `AGENTS.md`
> plus the Memora nodes under `root/skill/<name>` — every skill's own branch is
> its record, so there is no separate constitution file to consult.

## Choosing a part

| Situation | Part |
|---|---|
| Create a new skill from scratch, with evals and benchmarking | `parts/skill-creator/` (328K — read the sections you need) |
| Validate / score / grade a skill (structure, scripts, quality tiers) | `parts/skill-tester/` (302K — sectioned) |
| The authoring spec: frontmatter, structure, router shape, quality checklist | `parts/skill-authoring/` |
| Install or update a community skill from a local folder or archive | `parts/skill-installer/` |
| Local MCP ecosystem map (BrowseMCP, SearchMCP) | `parts/skill-hub/` |
| Scaffold a plugin bundle with versioned manifest + trust review | `parts/plugin-creator/` |

## Rules

- **One router shape.** Every top-level router carries the same four blocks —
  frontmatter trigger, `Choosing a part`, `Rules`, `Related` — as specified in
  `parts/skill-authoring/` §3. Quality is the body's content, not an extra
  layout.
- **Every router names two things in `Rules`**: the forks that go to
  `ask_questions`, and the skill's own Memora branch (`root/skill/<name>`) that
  it reads and appends to.
- **Creating a skill starts with the intake rule** in `parts/skill-authoring/`
  §0 (repeatability, no coverage in `skills/`, a place not a pile) —
  those parts are the how, not the whether.
- **After creating or editing a skill**: check it against
  `parts/skill-authoring/` §1–3 and run the skill-tester validation if scripts
  are involved. `_scripts/audit_skills_format.py` was retired on 2026-10-17, so
  the check is the spec plus reading the file.
- **Triggers must survive**: a description that does not match the body never
  fires — the August translation incident. Write descriptions in English, in
  the third person, with the words a real request would use.
- Skills live in `skills/`, are synced to Freebuff account-skills by
  `_scripts/sync_skills.py`; archives never delete.

## Related

`planning`, `review` (`verify`), `dev` (`using-superpowers`), `memory`.

From youlianvr/oper-share, CC-BY-4.0.

More from youlianvr/oper-share