Agent Skills
Agent skills are reusable, task-specific modules that bundle instructions and scripts for reliable agent execution. Each skill lives in .agent/skills/ and contains a SKILL.md file describing how to use it.
Directory structure
Each skill folder follows this baseline layout:
.agent/skills/<skill-name>/
SKILL.md
scripts/
...
references/
...
SKILL.md is required. The scripts/ and references/ directories are optional but recommended for repeatable tasks.
Using a skill
- Read the
SKILL.mdfile first. - Follow any ordered steps listed in the Usage section.
- Run the provided scripts from the repository root.
Example (UI standards audit):
bun .agent/skills/ui-standards/scripts/audit.mjs
Versioning & maintenance
- Keep skill instructions concise and scoped to a single outcome.
- Prefer scripts for repeatable checks or audits.
- Update
docs/agents/skills.mdwhen new skills are added so agents can find them quickly.
Current skills
- ui-standards: audits accessibility and UI quality checks via ESLint + Playwright.
- research-synthesis: converts external research into repo-scoped recommendations with confidence and verification checkpoints.
- verification-gate: produces a pre-commit verification checklist with scoped checks, broader gates, and open-risk notes.
- homepage-ui-polish: standardizes frequent homepage UI refinement work across hero/work/contact sections with accessibility-first validation.
- content-copy-refresh: streamlines recurring content/tone updates in markdown and metadata while preserving routes and discoverability.
- agent-doc-hygiene: keeps AGENTS docs, workflow references, and task-tracking files synchronized and drift-free.