By cfactolerin
Non-interactive multi-agent local-branch review run before pushing. Agents review the branch diff against base, arbiter synthesises a strict FINAL_REVIEW_RESULTS.md the implementing agent can act on.
Use this agent when dispatched by the code-reviewer:start skill to synthesise multiple agent reviews into a single FINAL_REVIEW_RESULTS.md. Reads all reviews, cross-examines if rounds allow, and produces the final strict report. NOT for direct user invocation.
Use this agent when dispatched by the code-reviewer:start skill to perform an independent local-branch code review. Reads the review prompt and current repo to produce a structured review. NOT for direct user invocation.
Use this agent when dispatched by the code-reviewer:start skill to run Codex CLI for an independent local-branch review. Shells out to the codex CLI and captures output. NOT for direct user invocation.
Use this agent when dispatched by the code-reviewer:start skill to run Gemini CLI for an independent local-branch review. Shells out to the gemini CLI and captures output. NOT for direct user invocation.
Use this agent when dispatched by the code-reviewer:start skill to run opencode CLI for an independent local-branch review. Shells out to the opencode CLI and captures output. NOT for direct user invocation.
Remove a review agent from the active agent list.
Add a review agent to the active agent list. Supported agents are claude, codex, gemini, and opencode.
Toggle the automatic `git push` review gate (config.auto_trigger).
Add, remove, or list dismissed findings in the current branch's DISMISSALS.md.
First-time setup wizard for the code-reviewer plugin. Configures agent list, agent settings, and Jira/Confluence credentials at ~/.code-reviewer/config.json.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
code-reviewer is a Claude Code plugin that runs a non-interactive,
multi-agent code review on the current local branch before you push it to
GitHub. Where prr is for the reviewer (after a PR is opened),
code-reviewer is for the author (right before they push): make the branch
ready for human review by catching bugs, security issues, regressions, and
missing tests automatically first.
Claude, Codex, Gemini, and opencode each review the branch diff against the
base branch in parallel. A Claude-powered arbiter cross-examines disagreements
and produces a single FINAL_REVIEW_RESULTS.md the implementing agent (and
you) can act on directly.
Required:
git — diff, log, merge-basejq — JSON config parsingpython3 — Jira fetch and HTML→Markdown (stdlib only; no pip install)Optional (enable additional reviewers):
codex agentgemini agent (uses Vertex AI; needs google_cloud_project / google_cloud_location)opencode agent. Requires OPENAI_API_KEY in your shell or opencode auth. The plugin does not store this key.Project-side linters are auto-detected and run if present:
rubocop (uses .rubocop.yml if present)ruff, mypy, flake8eslint, tscgolangci-lint (or go vet)cargo clippyshellcheckPlus tests: if the repo has a README.md, CLAUDE.md, AGENTS.md, or
CONTRIBUTING.md with a test/build/spec section, reviewers extract those
instructions and run the suite, flagging any regressions.
Inside a Claude Code session:
/plugin marketplace add cfactolerin/code_reviewer
/plugin install code-reviewer@cfactolerin-code-reviewer
Run once to configure agents and (optionally) Jira credentials:
/code-reviewer:setup
Then, when you have a branch ready to push:
/code-reviewer:start
Optional flags:
/code-reviewer:start --ticket ABC-123 # override Jira key
/code-reviewer:start --base origin/develop # override base branch
/code-reviewer:start --delta # fast review: only material new since last review
/code-reviewer:start --full # re-download Jira/Confluence; review entire branch
/code-reviewer:start --no-prune # keep all old round dirs (skip the 10-round prune)
/code-reviewer:start --ticket ABC-123 --base origin/release/v2
The skill runs all configured agents in parallel, the arbiter synthesises, and a final report is written to:
<review_output_path>/<repo-slug>/<branch-slug>/<YYYYMMDD-HHMMSS>/FINAL_REVIEW_RESULTS.md
(review_output_path defaults to /tmp/code-reviewer/ — see "Volatility note" below.)
After the review finishes, the skill hands back to the main Claude Code session with a prompt:
The code review found N findings (X critical, Y high, …). How would you like to proceed?
- Fix everything — apply fixes finding-by-finding, re-run for regression check
- Fix a subset — name the finding numbers to fix
- Discuss first — Q&A about findings before deciding
- Skip for now — leave the report and continue
Pick Fix everything for the typical case: the main session reads
FINAL_REVIEW_RESULTS.md, applies fixes one by one (running each finding's
acceptance check), then re-runs /code-reviewer:start. The previous report
is automatically included in the next review's context, so the arbiter
explicitly calls out regressions, items now fixed, and any newly introduced
issues.
| Skill | Description |
|---|---|
/code-reviewer:setup | First-time setup — agent list, Jira credentials, GCP project/location |
/code-reviewer:start [--ticket K] [--base R] [--delta|--full] [--no-prune] | Non-interactive multi-agent review of the current branch |
/code-reviewer:autodetect [true|false] | Toggle the git push review gate (config.auto_trigger). No-arg form prints the current state |
/code-reviewer:dismiss <fingerprint> | Add a finding fingerprint to DISMISSALS.md so it is excluded from future gate checks |
/code-reviewer:add-agent <name> | Enable an agent in your config (claude, codex, gemini, opencode) |
/code-reviewer:delete-agent <name> | Disable an agent in your config |
npx claudepluginhub cfactolerin/code_reviewer --plugin code-reviewerAI-powered PR review with parallel multi-agent review and arbiter synthesis
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.