GitHub Copilot Skills

Foundation ships a set of reusable GitHub Copilot Skills — structured prompt files that teach Copilot (and other AI agents, including Claude Code — the primary coding agent) how to perform common, multi-step development tasks consistently across every project that follows this architecture.

Skills live in .github/skills/<skill-name>/SKILL.md. They are picked up automatically by GitHub Copilot when the agent reads the repository.

Why Skills?

Without a skill	With a skill
Agent guesses the right lint/test commands	Agent runs the exact commands from `lefthook.yml`
Inconsistent fix strategies across sessions	Deterministic, documented procedure every time
Silent failures when a step is skipped	Required steps are explicit; status table always reported

Available Skills

Skill	Trigger phrase	Purpose
`pre-commit-check`	"run pre-commit checks", "fix lint errors", "green-light this branch"	Mirrors lefthook hooks; auto-fixes what it can; reports remaining issues

pre-commit-check

File: .github/skills/pre-commit-check/SKILL.md

📥 Download template

What it does

Mirrors the lefthook hooks exactly, then auto-fixes anything fixable, then reports remaining issues that require manual attention.

Hook	Tool	Scope	Auto-fixable?
pre-commit	BiomeJS `check --write`	`ui/` TS/TSX/JS/CSS	✅ Yes
pre-commit	Ruff `check --fix` + `format`	`api/` Python	✅ Yes
—	TypeScript `tsc --noEmit`	`ui/web-app`, `ui/marketing`	❌ Manual fix required
pre-push	pytest	`backend/api`	❌ Manual fix required
pre-push	Vitest	`ui/web-app`	❌ Manual fix required

When to invoke it

Use this skill before committing, before opening a PR, or any time you want to confirm a branch is clean:

"Run pre-commit checks"
"Fix lint errors"
"Fix type errors"
"Green-light this branch"
"Pre-push check"

Procedure summary

Identify scope — check git status for changed packages; honor any argument passed (e.g. "api").
BiomeJS auto-fix — pnpm biome check --write . from the repo root, then verify with pnpm biome check ..
Ruff auto-fix — uv run ruff check --fix . && uv run ruff format ., then verify.
TypeScript type-check (web-app) — cd ui/web-app && pnpm exec tsc --noEmit. Fix errors manually.
TypeScript type-check (marketing) — cd ui/marketing && pnpm exec tsc --noEmit. Fix errors manually.
pytest — cd backend/api && uv run --extra dev pytest. Never delete tests to make them pass.
Vitest — cd ui/web-app && pnpm test.
Final verification — re-run all checks; every command must exit with code 0.

Output

The skill always ends with a status table:

Check	Status	Notes
BiomeJS lint/format	✅ / ❌	Remaining manual-fix items
Ruff lint/format	✅ / ❌	Remaining manual-fix items
TypeScript — web-app	✅ / ❌	Error count
TypeScript — marketing	✅ / ❌	Error count
pytest	✅ / ❌	Passed / total
Vitest	✅ / ❌	Passed / total

Adding a New Skill

Create .github/skills/<skill-name>/SKILL.md in the target repository.

Add the required YAML front matter:

---
name: <skill-name>
description: '<one-line description + USE FOR: trigger phrases>'
argument-hint: 'Optional: ...'
---

Write the procedure using numbered steps with exact shell commands.
End with a reporting section (status table or structured output).
Copy the file to public/templates/skills/<skill-name>/SKILL.md in this repo and add a row to the Available Skills table above.

Skills are plain Markdown. They do not require special tooling — any AI agent that reads the repository can follow them.