Blog Post · Architecture Deep Dive
Adversarial multi-agent pipeline for Claude Code. Separate AI agents generate and critique each other's work in GAN-style loops — generators produce artifacts, discriminators validate them, feedback drives convergence. Each agent runs in its own context window with fresh perspective.
Install
# Per-project (teammates get it automatically) cp -r .claude/skills/ /path/to/your-project/.claude/skills/ # Personal (all projects) cp -r .claude/skills/ ~/.claude/skills/
Requires Claude Code CLI and a git-initialized project.
Skills
| Skill | Purpose | Output | Next Step |
|---|---|---|---|
/brainstorm |
Interactive design session — explores codebase, asks scoping questions | brainstorm.md |
/pipeline |
/audit |
Combined audit runner — select any combination of eval, health, docs | Multiple intake docs | /pipeline |
/repo-eval |
3-evaluator panel scoring 12 pillars | eval.md |
/pipeline |
/repo-health |
Technical debt audit across 4 vectors | health-audit.md |
/pipeline |
/doc-health |
Documentation drift detection across 6 phases | doc-audit.md |
/pipeline |
/pipeline |
Automated build/remediation cycle — routes by intake doc type | Committed code | Done |
Usage
# Feature development /brainstorm I want to add webhook support for payment events /pipeline 2026-03-12-payment-webhooks # Full audit (health → eval → docs) with one pipeline run /audit all /pipeline 2026-03-15-audit-my-app # Or run individual audits (each creates its own plan directory) /repo-eval /pipeline 2026-03-15-eval-my-app
Resume any interrupted pipeline by re-running /pipeline with the same slug.
Pipeline Flows
Feature (brainstorm.md)
Planner ↔ Plan Reviewer → Implementer ↔ Code Reviewer → Final Reviewer
(max 3 iter) (max 3 iter/phase) GO/NO-GO
Repo Eval (eval.md)
3 Evaluators → Planner ↔ Plan Reviewer → Implementer ↔ Reviewer → Verify
(parallel) (max 3) (max 3/phase) verify findings
Repo Health (health-audit.md)
Auditor → Planner ↔ Plan Reviewer → Hygienist ↔ Health Reviewer → Fortifier ↔ Health Reviewer → Verify
[cleanup] [guardrails] verify findings
Doc Health (doc-audit.md)
Doc Auditor → Planner ↔ Plan Reviewer → Doc Engineer ↔ Doc Reviewer → Verify
[fix + prevent] verify findings
File Structure
.claude/skills/
├── audit/SKILL.md # Combined audit runner
├── brainstorm/SKILL.md
├── repo-eval/SKILL.md
├── repo-health/SKILL.md
├── doc-health/SKILL.md
└── pipeline/
├── SKILL.md # Orchestrator (routes by intake doc type)
├── pipeline-protocol.md # Signal protocol spec
├── planner.md # Shared across all flows
├── plan_reviewer.md # Shared across all flows
├── implementer.md # Feature + repo-eval flows
├── reviewer.md # Feature + repo-eval flows
├── final_reviewer.md # Feature flow only
├── eval-hire.md # The Pragmatist
├── eval-stress.md # The Oncall Engineer
├── eval-day2.md # The Team Lead
├── health-auditor.md # Pure assessment, no fix guidance
├── health-hygienist.md # Subtractive (delete, simplify)
├── health-fortifier.md # Additive (lint, CI, hooks)
├── health-reviewer.md # Reviews both hygienist + fortifier
├── doc-auditor.md # 6-phase drift detection
├── doc-engineer.md # Fix docs + add prevention
├── doc-reviewer.md # Reviews doc changes
└── flows/
├── audit-flow.md # Unified plan across multiple audit types
├── repo-eval-flow.md
├── repo-health-flow.md
└── doc-health-flow.md
License
MIT — see LICENSE.