4-agent code review. Correctness, security, performance, style. Auto-fixes NITs.
From godmodenpx claudepluginhub arbazkhan971/godmodeThis skill uses the workspace's default tool permissions.
Designs and optimizes AI agent action spaces, tool definitions, observation formats, error recovery, and context for higher task completion rates.
Enables AI agents to execute x402 payments with per-task budgets, spending controls, and non-custodial wallets via MCP tools. Use when agents pay for APIs, services, or other agents.
Compares coding agents like Claude Code and Aider on custom YAML-defined codebase tasks using git worktrees, measuring pass rate, cost, time, and consistency.
/godmode:review, "review this", "check my code"git diff main...HEAD
Read every changed file. >500 lines -> split into per-directory reviews.
Each agent outputs:
SEVERITY | FILE:LINE | DESCRIPTION | SUGGESTED FIX
Severities: MUST-FIX, SHOULD-FIX, NIT.
IF no findings: state {category}: No issues found.
Combine findings. Same file:line = merge, keep highest severity, first agent's fix. Sort: MUST-FIX first.
SAFE: imports, formatting, whitespace, comment typos.
UNSAFE (human): logic, API, security, test changes.
Each auto-fix = separate commit:
"review-autofix: {description}".
Score: 0-10 (median of all agents).
IF score < 8: author responds to every MUST + SHOULD. Re-review only modified parts. Max 3 iterations.
Append .godmode/review-log.tsv:
timestamp scope category severity file_line description status
KEEP if: finding has file:line + severity + fix.
DISCARD if: vague, lacks evidence, or duplicate.
On finding DISCARD: classify and append to .godmode/review-failures.tsv.
False positive classes: stale_context, framework_idiom, test_only_code, intentional_pattern, duplicate_finding.
For higher-confidence reviews, dispatch agents with different models:
Findings caught by only one model are flagged as "cross-model finding" — these are the blind spots that self-play misses.
Enable with: --cross-model flag or when reviewing security-critical code.
STOP when FIRST of:
- All 4 agents completed and findings merged
- Max 3 re-review iterations
- Re-review found 0 new issues
On failure: git reset --hard HEAD~1. Never pause.
| Failure | Action |
|---|---|
| Diff too large (>500) | Split per-directory reviews |
| Vague findings | Discard, re-run with explicit instruction |
| Auto-fix breaks tests | Revert, reclassify as SHOULD-FIX |
| Conflicting findings | Escalate to highest severity |