Self-verify work, run review agents in parallel, fix P1s, report findings
Self-verifies code changes, runs parallel security and quality reviews, and fixes critical issues before deployment.
/plugin marketplace add https://www.claudepluginhub.com/api/plugins/brite-nites-britenites-plugins-britenites/marketplace.json/plugin install brite-nites-britenites-plugins-britenites@cpd-brite-nites-britenites-plugins-britenitesYou are reviewing work before it ships. Your job is to verify correctness, run specialized review agents, fix critical issues, and produce a clean report for the developer.
Before running review agents, confirm the Task tool works:
This catches the case where you'd wait for 3 parallel agents that all silently fail.
Before launching review agents, verify your own work against the execution plan:
package.json scripts, CLAUDE.md, or common patterns like npm test, npx vitest, npx jest). If no test suite exists, note it and proceed.npm run build (or equivalent) to catch type errors and build failures.git diff and read through every change. Look for:
If self-verification reveals issues, fix them before proceeding to agent review.
Dispatch three specialized review agents in parallel using the Task tool. Each agent reviews the current diff against the codebase.
Prepare the review context first:
git diff main...HEAD (or appropriate base branch) to capture all changesLaunch all three simultaneously:
code-reviewer agent — "Review the code changes on this branch for bugs, logic errors, and quality issues. The diff is from git diff main...HEAD. Use P1/P2/P3 severity."
security-reviewer agent — "Review the code changes on this branch for security vulnerabilities. The diff is from git diff main...HEAD. Use P1/P2/P3 severity."
typescript-reviewer agent — "Review the code changes on this branch for TypeScript, React, and Next.js issues. The diff is from git diff main...HEAD. Use P1/P2/P3 severity."
Wait for all three to complete.
Merge findings from all three agents into a single report, deduplicated and sorted by severity:
## Review Findings
### P1 — Must Fix
- [Finding from agent] — [file:line]
- ...
### P2 — Should Fix
- [Finding from agent] — [file:line]
- ...
### P3 — Nit
- [Finding from agent] — [file:line]
- ...
---
**Totals**: X P1, Y P2, Z P3
**Sources**: code-reviewer (A findings), security-reviewer (B findings), typescript-reviewer (C findings)
If there are P1 findings:
If there are no P1 findings, skip to Step 5.
Present the final state to the developer:
## Review Complete
**P1 (fixed)**: [list what was fixed, or "None"]
**P2 (your call)**: [list remaining P2s with context]
**P3 (FYI)**: [list P3s briefly]
**Tests**: Passing / Failing (details)
**Build**: Clean / Errors (details)
**Verdict**: Ready to ship / Needs your input on P2s / Blocked on P1
If all P1s are fixed and tests pass, suggest: "Ready for /britenites:ship when you are."
If P2s need decisions, ask the developer which to fix and which to accept.