AI Skills & Agentstjcageslinear-discipline
linear-discipline
Keep Linear tracking honest during development with search-before-create, lifecycle updates, and close-the-loop discipline.
Instalação
npx @compound-design/skills get tjcages/linear-disciplineConteúdo
Linear discipline (always-on)
Chat rules: RESPONSE.md. Auth before Linear calls: AUTH.md.
Detect tracking
Tracked if the repo has a “Linear tracking” (or equivalent) section in CLAUDE.md, AGENTS.md, or .cursor/rules/* that names team + project.
Skipped if .linear-tracking-skip exists at repo root.
If tracked
Follow the protocol written in-repo:
- Search before create.
- Non-trivial work → file/update an issue in the named project + milestone.
Backlog→In Progressat start →Doneonly when shipped.- Close the loop before session end (state and/or comment).
- Prefer Linear’s generated branch names.
Before any Linear tool call: Step 0 auth.
If untracked and not skipped (soft nudge — 1B)
Ask once, one line, no preamble:
This repo isn’t tracked in Linear. Set it up now? (~10–20 min)
- Yes → hand off to
linear-setup(do not bootstrap inline without that skill’s gates). - No → create
.linear-tracking-skipat repo root; stay quiet next time. - Never auto-create Linear projects or issues from a nudge alone.
- Never nudge from Automation /
linear-monitorruns.
If skipped
Stay quiet about setup. Still use Linear MCP for explicit user requests.
Out of scope
Full bootstrap → linear-setup. Rescue/audit → linear-sync. Health digest → linear-monitor.
Relacionadas
Escrito por tjcages em tjcages/linear, sob licença MIT. Origem: https://github.com/tjcages/linear/blob/main/skills/linear-discipline/SKILL.md