General-purpose coding policy for Baruch's AI agents
95
91%
Does it follow best practices?
Impact
96%
1.31xAverage score across 10 eval scenarios
Advisory
Suggest reviewing before use
Reference for Step 7 of the install-reviewer skill. Pulled out of SKILL.md so the main flow stays scannable.
The PR body the skill opens must include the following content blocks. Order them as below.
Explain that the two workflows install jbaruch/coding-policy at run time and review every PR against it. The OpenAI and Anthropic reviewers each self-gate on the PR's Author-Model: declaration so the active reviewer is cross-family whenever the declaration permits. When the declaration spans both paired families, or neither paired family, both reviewers run as the documented fallback. See jbaruch/coding-policy: author-model-declaration on the registry.
List the three repository secrets the consumer must set BEFORE merging:
OPENAI_API_KEY — OpenAI billing account for the Codex reviewerANTHROPIC_API_KEY — Anthropic billing account for the Claude Code reviewerTESSL_TOKEN — generated at https://tessl.io/account/api-keys, used by the workflow's tessl install setup stepAdd a one-line note that merging without all three secrets set will cause the workflows to fail on their first run.
Note that each reviewer's verdict begins with a one-line load indicator:
Policy loaded: N rule files from /tmp/gh-aw/coding-policy/.tessl/tiles/jbaruch/coding-policy/rules/ (installed tile).Tells the consumer at a glance whether the workflow setup tessl install ran cleanly or whether the policy never reached the runtime.
If Step 1's preflight warnings array is non-empty, add an ## Action required before merge section that reproduces each warning's reason verbatim. These are advisory findings the install-reviewer skill deliberately does NOT auto-fix — the section exists so the consumer sees and acts on the finding instead of discovering it via a failed workflow.
If Step 1 emitted no warnings, omit the section entirely.