General-purpose coding policy for Baruch's AI agents
91%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
A script that runs Step 2 or Step 7 unattended (release.sh, merge-and-cleanup.sh) keeps every gate the interactive flow keeps. It calls the same scripts in the same order, gates on their exit codes, and fails loudly on any unmet gate.
<type>(<scope>): <imperative summary>; never pass a raw title straight to gh pr create.ready. The watcher owns readiness; the wrapper never duplicates its conditions.review_bodies[].body or non-empty inline_comment_bodies[].body is unread, from every author, page and state. Resume after an agent or human reads every body and blocking findings are addressed through Step 6. The release agent may be the reader. Ordinary advisory replies and thread resolution stay outside the merge predicate. Counts, states and displayed text never satisfy this gate.git branch -d, never -D.git rev-parse --git-dir differs from --git-common-dir), follow Step 7 path (B) in order.skills/release/references/PUBLICATION.md. Report the release confirmed only when every channel confirms..tessl-plugin
rules
skills
coding-contracts
references
migrate-to-plugin
onboard-repo
release
hooks
references