Add /review-loop command + orchestrating-agent-relay schema/loop docs#50
Add /review-loop command + orchestrating-agent-relay schema/loop docs#50khaliqgant wants to merge 5 commits into
Conversation
Adds the review-loop slash command: Claude and Codex each review the working-tree diff independently while fanning out into their own subagents, a Codex fixer repairs every valid finding from both reviews and hardens it with tests, and the loop repeats with fresh agent context until both reviewers independently sign off on the same post-fix state. - commands/review-loop.md: harness-agnostic command spec - prpm.json: register review-loop as a slash-command package Pairs with the workflow at AgentWorkforce/workflows repeatable/review-loop/workflow.ts. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Require pinning the fetched workflow to an immutable ref (published package version, git tag/release, or commit SHA); never fall back to the default branch silently. (CodeRabbit: deterministic-run concern) - Add `bash` language to the run-loop fenced block (markdownlint MD040). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- commands/review-loop.md: add step verifying the agent-relay CLI is available before dry-run/launch; use full raw.githubusercontent URL for the commit-SHA example (cubic PR #49 P3) - commands/create-workflow.md: same agent-relay CLI verification step - writing-agent-relay-workflows: fix implementation-reconcile race by depending on the impl agent steps (failOnError:false) instead of a shared 'context' dep; restore pr_url to the valid verification-types row in the Common Mistakes table (relay PR #912 feedback) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- writing-agent-relay-workflows 1.6.12 -> 1.6.13 (reconcile race + pr_url table fix) - create-workflow 1.0.2 -> 1.0.3 (agent-relay CLI precheck step) - review-loop 1.0.0 -> 1.0.1 (agent-relay CLI precheck step + full raw.githubusercontent URL; also covers the un-bumped aadeff5 docs edit) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…estrating-agent-relay Address agent feedback on the orchestrating-agent-relay skill: - Document the verified replies --json schema (id/from/to/text/createdAt/ direction); call out createdAt (not ts) and the hard-coded inbound direction, verified against relay CLI source - Note the no-conversation empty state is a plain string, not [] - Add Multi-Round Review Loops section: workers must not self-remove until told, respawn-with-full-context fallback, silent-death detection - Expand Common Mistakes table and orchestrator Protocol accordingly - Bump orchestrating-agent-relay 2.0.0 -> 2.1.0 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (5)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Superseded by a clean branch rebased on main (the /review-loop commits were already merged in #49, polluting this PR's diff). |
Summary
This branch bundles two related pieces of work around the multi-agent review loop:
/review-loopdual-reviewer slash command (existing commits) — adds the command, addresses CodeRabbit feedback, and includes an agent-relay CLI precheck fix.orchestrating-agent-relayskill updates (new) — addresses agent feedback from running a real review→fix→re-review loop.orchestrating-agent-relay changes
Driven by concrete feedback from an orchestrator agent that hit several footguns:
replies --jsonschema is now documented and source-verified againstrelay/src/cli/commands/messaging.ts:id,from,to,text,createdAt,direction(+ optionalunread/unread_state)createdAt, nottsreplies --json,directionis always the hard-coded literal"inbound"(never"incoming"/"from"/"outbound");"outbound"only appears inhistory --to <agent> --jsonNo DM conversation with <Name>., not[]— with a copy-paste guardwho --jsonliveness + wall-clock fallback, since DM monitors only fire on DMs)orchestrating-agent-relay2.0.0 → 2.1.0🤖 Generated with Claude Code