Repository navigation
playbook 2.2.0
Version 2.2.0 brings the skills in line with the best practices for skills that Anthropic and agentskills.io publish, and breaks nothing.
What's new
- A stricter way to write a skill.
write-skilland its style guide add rules for a skill's name, reference files one level deep, an optional Gotchas section for mistakes seen, writing from the project's own material, and scripts that print their usage, write messages to standard error, and exit with one documented code per kind of failure. - Cases of behaviour. A new skill now gets three cases that compare the agent with the skill and without it, written before its
SKILL.md, besides its two cases of triggering. The product's skills have theirs, under each plugin'sevals/. - Seven skills reviewed against the new rules: shorter, without rules repeated from the contract or the working rules, and with the same jobs.
/playbook:initrecords a choice about a skill of your project only in the open work; the wiki's page of skills changes when that work closes.- The doctor now moves "Where work stands" under "Phases before format 4" as a heading the wiki's check accepts.
- The scripts of
/playbook:initand of the doctor answer--help.
Update
claude plugin marketplace update playbook
claude plugin update playbook@playbookThen update each of playbook-wiki@playbook and playbook-agents@playbook that claude plugin list shows. The changes work from the next session.
Licence
MIT. Issues and pull requests are welcome; see CONTRIBUTING.md.