fight

/fight

Run every fight-club reviewer against the target and produce a combined report.

The user invoked this command with: $ARGUMENTS

Step 1: Get the code

Parse $ARGUMENTS to determine the target:

• File path → read the file
• Number → gh pr diff <number> to get the PR diff
• "diff" or empty → git diff HEAD

Step 2: Run all reviewers

Apply each of these lenses to the code. For each, produce findings using that skill's output format:

1. Adversarial Architect — design, factoring, coupling, abstraction
2. Adversarial Auditor — vulnerabilities, trust boundaries, data exposure
3. Adversarial Optimizer — scaling assumptions, inefficiencies, hot paths
4. Adversarial QA — coverage gaps, weak assertions, missing failure modes
5. Adversarial SRE — production failure scenarios, silent failures, recovery risk

Step 3: Aggregate

After all reviewers, produce a summary:

## Fight Club Report

### Adversarial Architect
[findings or "No issues found"]
Verdict: Composable | Acceptable | Poor | Broken

### Adversarial Auditor
[findings or "No issues found"]
Verdict: Secure | Hardening Needed | Vulnerable | Critical

### Adversarial Optimizer
[findings or "No issues found"]
Verdict: Efficient | Optimization Needed | Will Not Scale | Blocking Issue

### Adversarial QA
[findings or "No issues found"]
Verdict: Solid | Needs Work | Insufficient | No Confidence

### Adversarial SRE
[findings or "No issues found"]
Verdict: Ready to Ship | Needs Hardening | High Risk | Do Not Ship

---

## Blocking (must fix before merge)
[Cross-lens list of every finding with Blocking: Yes. Empty section if none.]

## Follow-up (track as issues, do not block merge)
[Cross-lens list of every finding with Blocking: No. Group by lens.]

---
## Overall Verdict
[APPROVED / APPROVED (with follow-ups) / NEEDS WORK / REJECT]
[2-3 sentences: biggest risks, blockers, recommended next action]

Verdict rules

The overall verdict is determined by Blocking findings, not by severity alone: