Claude

What gets rendered

# repo-owned context can live above the import
@AGENTS.md

The @AGENTS.md import pulls the full rendered standard — core rules plus the repo's profile stacks — into every Claude session. Scoping is inherited from AGENTS.md's per-section prose.

Connect Claude Code (CLI / IDE)

1. Onboard the repo with npx redlinegate init — the pull request it opens adds CLAUDE.md and AGENTS.md.
2. Nothing else. Claude Code loads CLAUDE.md automatically on session start; reviews and edits follow the standard, including the output contract when asked to review.

Connect Claude in GitHub

1. Install the Claude GitHub app on the org or repo.
2. Mention @claude on a PR, or configure it to review automatically.
3. Findings arrive with the Redline/<SEVERITY> [<rule-id>]: prefix, so telemetry attributes them like any other vendor.

Compare it against other vendors

$ npx redlinegate metrics score-seeds --repo acme/pilot-claude  --pr 7 --json
$ npx redlinegate metrics score-seeds --repo acme/pilot-copilot --pr 4 --json
# compare blocker_recall and false_positives_on_clean on identical seeded input

The scorer reads the contract prefix and a bot list — it neither knows nor cares which vendor produced the comment.