Job pack · harness
Skill pack
Progressive-disclosure expertise. Write descriptions that trigger. Distill from traces. Do not flood the constitution.
When: You are packaging how an agent should do a job for a repo or a team.
Description first (what + when). Body as a playbook. references/ on demand. Trigger-test five phrasings.
GET /api/canon/jobs/skill-pack?format=md
Playbooks
- Write a skillFrontmatter that triggers, a body that runs, files on demand. Distill from traces if you have them.
- Write AGENTS.mdThe constitution of the repo. Commands, boundaries, done. Point at skills for depth.
- Distill a traceCompile a paid-for lesson into a skill, a rubric, or a golden case. Do not hoard transcripts.
Doctrine to load
- Writing Skills That TriggerA skill that does not trigger is documentation nobody opened. The description is a search query the model will match against the user's request. Write it as if under-triggering were the only bug — because it is.
- Skills and Progressive DisclosureA skill is a folder with a SKILL.md. Metadata always sits in context; the body loads on trigger; scripts and references load on demand. This is how you scale expertise without flooding the window.
- AGENTS.md as ConstitutionAGENTS.md (and CLAUDE.md, and the Cursor rules file) is the repo's standing prompt. It is owned policy: how to build, test, and stop. Keep it short, imperative, and true, or the agent will ignore it and invent a culture.
- Distill Traces into SkillsA trace is a lesson the agent already paid for. Compiling it into a skill, a rubric, or a golden eval is how the system gets cheaper and less lost. Uncompiled traces are landfill.
Refuse
- Unowned prompt. The policy lives in a framework black box you cannot diff, test, or revert.
- Context hoarding. Raw HTML, full CI logs, six PDFs, entire JSON envelopes stuffed 'just in case'. The goal is the first thing forgotten.
- Memory as unfiltered logs. Session transcripts stuffed into the next session. The user cannot inspect or delete them.
Checklists
Skill quality
- name + description say what it does and when to use it, with trigger contexts.
- SKILL.md under ~500 lines. Depth in references/, deterministic work in scripts/.
- Imperative voice. Explain why. No ALWAYS/NEVER wallpaper.
- One job. No overlap that makes the model pick none.
- Does not duplicate AGENTS.md. The constitution points; the skill teaches.
Recipes
Default corpus
anthropic-skillsaddy-agent-skillsopenai-skillscontext-engineering-skills