Installs into .claude/skills of the current project.
Are you the author of Writing Skills?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/charlieviettq-writing-skills)
---
name: writing-skills
description: "Author and pressure-test agent skills—triggers, thin SKILL.md, references, validation. Use when creating or revising skills in this repo."
allowed-tools: Read, Glob, Grep
---
# Writing skills
## When to use
- Adding a skill under `.cursor/skills/<domain>/<name>/`
- Fixing discovery issues (agent never invokes the skill)
- Splitting an oversized skill into SKILL.md + reference.md
## When not to use
- One-off project instructions (use README or rules)
- Copying entire third-party skill trees without audit
## Workflow
1. Read [`docs/skill-writing-guide.md`](../../../../docs/skill-writing-guide.md) and [`docs/review-rubric.md`](../../../../docs/review-rubric.md)
2. Run `skill-supply-chain-audit` if source is external
3. Draft frontmatter: `name` matches folder; description includes **Triggers:**
4. Keep SKILL.md thin; move long examples to `reference.md`
5. Pressure-test: list 3 user phrases that should trigger and 2 that should not
6. Run `python3 scripts/validate-skills.py`
7. Regenerate Claude skills if editing Cursor source
## Description quality
- **Good:** concrete workflow + quoted trigger phrases
- **Bad:** broad "use for everything" or "MUST run before any action"
## Checklist before PR
- [ ] Public-safe (no secrets, no auto-push scripts in skill body)
- [ ] No duplicate of existing skill in `SKILL_INVENTORY.md`
- [ ] Entry in `EXTERNAL_SKILLS.md` if adapted from upstream
- [ ] `CHANGELOG.md` updated under Unreleased
## Related
`skill-supply-chain-audit`, `reflect-yourself`, `core-workflow/planning-and-task-breakdown`
*Authoring practices inspired by [obra/superpowers](https://github.com/obra/superpowers) (MIT).*
