Skip to main content

Skills

Run a skill by typing its name, for example /implement TOOL-123.

SkillWhat it does
/ticketCreate, read or update a Linear ticket
/epicCreate a Linear project with its tickets
/readGroom a ticket: explore the code, estimate size, draft the technical plan
/implementAssign a ticket, branch from origin/main, and plan the work
/commitCommit with the ticket ID taken from the branch name
/reviewRun Biome, the type check and tests, then review the diff
/prPush and open a draft PR from the repo's template
/qaTest the branch's changes in a browser with Playwright
/fixRun a bug fix end to end: read, implement, commit, QA, PR
/triageInvestigate a bug with its Slack thread and post the findings

Planning​

/ticket  β†’  /read  β†’  /epic (for larger work)

Ends with a groomed ticket in Linear: acceptance criteria, a technical plan and a size.

Implementation​

/implement  β†’  write the code  β†’  /commit  β†’  /review  β†’  /pr

Ends with a draft PR that has passed the repo's checks and a local review. Run /qa before /pr for anything visible in a browser.

Per-repo configuration​

Some skills read a file from the project repo, so their behavior differs between repos:

FileRead by
.claude/rules/architecture.md/read, /implement, /review (and they add to it)
.claude/rules/pr-review.md/review
.claude/rules/qa.md/qa, for URLs, login and routes
.github/pull_request_template.md/pr

Adding a skill​

  1. Add a folder under skills/ with a SKILL.md. Its frontmatter sets the name, description, allowed tools and argument hint.
  2. Refer to the skill's own files with relative paths (.claude/skills/<name>), so it works from any repo.
  3. If it has scripts, give it a package.json and run them with npx tsx.
  4. If it needs keys, add a .env.example.
  5. A skill that builds on others should call them, not copy their steps.