Skip to content

majiayu000/spellbook / skills/skill-creator

skill-creator

Verified

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update 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.

FileKindSize
LICENSE.txtmarkdown11.1 KB
SKILL.mdskill md33.1 KB
agents/analyzer.mdmarkdown10.1 KB
agents/comparator.mdmarkdown7.1 KB
agents/grader.mdmarkdown8.8 KB
assets/eval_review.htmlother6.9 KB
eval-viewer/generate_review.pyscript16.0 KB
eval-viewer/viewer.htmlother44.0 KB
references/quality-playbook.mdmarkdown2.9 KB
references/schemas.mdmarkdown13.0 KB
requirements.txtmanifest17 B
scripts/aggregate_benchmark.pyexecutablescript15.0 KB
scripts/generate_report.pyexecutablescript12.5 KB
scripts/improve_description.pyexecutablescript10.5 KB
scripts/package_skill.pyexecutablescript5.9 KB
scripts/path_safety.pyscript3.2 KB
scripts/quick_validate.pyexecutablescript5.8 KB
scripts/run_eval.pyexecutablescript11.9 KB
scripts/run_loop.pyexecutablescript13.4 KB
scripts/utils.pyscript1.6 KB

references/quality-playbook.md

View on GitHub
# Skill Quality Playbook

Use this reference when creating or materially improving a skill. Keep the main
`SKILL.md` short; put large reference tables, scripts, templates, and examples
behind progressive disclosure.

## Pick One Primary Type

Choose the skill's main job before writing:

- Library/API reference: how to use a library, CLI, SDK, or internal platform.
- Product verification: scripts or tool flows that prove product behavior.
- Data fetching/analysis: canonical queries, dashboards, metric names, helpers.
- Business process automation: repeatable team workflow with required state.
- Code scaffolding/templates: boilerplate plus natural-language requirements.
- Code quality/review: deterministic style, review, testing, or policy checks.
- CI/CD/deployment: build, smoke test, deploy, monitor, rollback, PR babysitting.
- Runbook: symptom-driven investigation that returns a structured report.
- Infrastructure operations: guarded maintenance, especially risky actions.

If a draft has two or three primary types, split it or make `SKILL.md` a router
that points to narrower references.

## High-Signal Sections

- **Description**: Write for triggering, not summary. Include when to use the
  skill, user phrasing, symptoms, and near-boundary cases.
- **Gotchas**: Add observed failure modes, naming mismatches, false-success
  signals, risky defaults, and recovery checks.
- **Verification**: For objective workflows, include scripts, assertions, smoke
  tests, browser checks, health checks, or explicit done-when proof.
- **Reliable Skill Contract**: For high-value workflow skills, include negative
  examples, phase checkpoints, machine-checkable done conditions, replay or
  smoke hooks, and drift signals.
- **Support files**: Use `references/`, `scripts/`, `assets/`, `templates/`, and
  `evals/` when detail would otherwise bloat `SKILL.md`.
- **Setup**: If the skill needs user or environment context, describe a config
  file or setup flow instead of asking repeatedly.
- **Memory**: Repeated workflows can keep append-only logs, JSON, or SQLite state
  when previous runs change the next run.
- **Hooks**: Use on-demand hooks only for temporary guardrails that would be too
  annoying to run globally.

## Authoring Checks

Before calling a skill good, ask:

1. Would the model know to trigger this from the `description` alone?
2. Does the skill teach something non-obvious or repo/org-specific?
3. Are the common failure modes captured as gotchas?
4. Does `SKILL.md` point clearly to every support directory it expects to use?
5. Is there a concrete verification path for objective work?
6. Are risky or destructive actions guarded by confirmation or checks?
7. Does a high-value workflow skill include the five Reliable Skill Contract
   elements or record why a specific element is deferred?

Run the repository-level audit when available:

```bash
python3 scripts/audit_skill_quality.py <skill-name>
```
finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the workStars: 322kUpdated Claude CodeCodexVerifiedInstall
systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixesStars: 322kUpdated Claude CodeCodexVerifiedInstall
test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation codeStars: 322kUpdated Claude CodeCodexVerifiedInstall
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: 198kUpdated Claude CodeCodexVerifiedInstall