CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

jbaruch/coding-policy

General-purpose coding policy for Baruch's AI agents

74

Quality

93%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

step-4-review.mdskills/release/references/

Step 4 — Reviews Fire Automatically

Opening, reopening or pushing to a same-repository PR starts both independent reviewers. Codex checks policy compliance. Claude reviews code quality and bugs with a replacement system prompt and no loaded policy, agent instructions, plugins, skills, hooks or memory. Both authenticate through subscriptions.

  • In coding-policy, review-codex.yml posts as github-actions[bot]; review-claude.yml posts as jbaruch-claude-reviewer[bot].
  • In consumers, review-trigger.yml dispatches fleet-review.yml; its policy and quality jobs post as coding-policy-fleet-reviewer[bot] and jbaruch-claude-reviewer[bot], respectively. The marker-gated scheduled poll is a backstop, deduplicated separately for each lane.

Policy and quality use distinct GitHub identities. An approval from either cannot supersede the other's change request in GitHub. Identity and provenance follow skills/release/references/REVIEW_DETAILS.md, “Claude — independent code-quality review.” Both reviewers classify findings as blocking or advisory under rules/review-severity.md. Blocking findings from either hold the merge; both bodies must arrive and be read.

Fork heads receive no credentials. Adopt them with Skill(skill: "adopt-fork-pr"). A push without an open PR does not trigger these reviewers. If a lane fails to arrive, inspect its workflow and authentication, repair the cause and rerun the failed job; do not request Copilot or invent a replacement polling loop.

Proceed immediately to Step 5.

README.md

tile.json