JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 160
  • Score
    100M100P100Q114994F
  • License MIT

Adversarial AI agent team for any project — installs Claude Code agents, hooks, and skills that enforce quality through productive friction

Package Exports

  • @fredericboyer/dev-team
  • @fredericboyer/dev-team/dist/init.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@fredericboyer/dev-team) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

dev-team

Adversarial AI agent team for any project. Installs Claude Code agents, hooks, and skills that enforce quality through productive friction.

Instead of an AI that agrees with everything, dev-team gives you six opinionated specialists that challenge each other — and you.

Install

npx @fredericboyer/dev-team init        # Interactive wizard
npx @fredericboyer/dev-team init --all  # Everything, no prompts

Requires Node.js 18+ and Claude Code.

What you get

Agents

Agent Role When to use
@dev-team-voss Backend Engineer API design, data modeling, system architecture, error handling
@dev-team-mori Frontend/UI Engineer Components, accessibility, UX patterns, state management
@dev-team-szabo Security Auditor Vulnerability review, auth flows, attack surface analysis
@dev-team-knuth Quality Auditor Coverage gaps, boundary conditions, correctness verification
@dev-team-beck Test Implementer Writing tests, TDD cycles, translating audit findings into test cases
@dev-team-deming Tooling Optimizer Linters, formatters, CI/CD, hooks, onboarding, automation

Szabo and Knuth use Opus (deep analysis, read-only). Voss, Mori, Beck, and Deming use Sonnet (implementation, full access).

Hooks

Hook Trigger What it does
Safety guard Before Bash commands Blocks dangerous commands (rm -rf, chmod 777, force push to main, DROP TABLE/DATABASE, curl|sh, and others)
TDD enforcement After file edits Blocks implementation changes when no corresponding test file exists and no tests were modified in the session
Post-change review After file edits Routes files to domain agents for review (security → Szabo, API → Mori, infra → Voss, tooling → Deming, code → Knuth). Advisory, does not block.
Pre-commit gate On task completion Reminds about running review agents before committing. Advisory, does not block.
Task loop On stop Manages iterative task loop with adversarial review gates

All hooks are Node.js scripts — work on macOS, Linux, and Windows.

Skills

Skill What it does
/dev-team:challenge Critically examine a proposal, approach, or implementation
/dev-team:task Start an iterative task loop with adversarial review gates

Challenge protocol

Agents challenge each other using classified findings:

  • [DEFECT] — Concretely wrong. Blocks progress.
  • [RISK] — Not wrong today, but creates a likely failure mode. Advisory.
  • [QUESTION] — Decision needs justification. Advisory.
  • [SUGGESTION] — Works, but here's a specific improvement. Advisory.

Rules: concrete evidence required, one exchange before escalation, human decides disputes.

How it works

Task loop (/dev-team:task)

For non-trivial work, the task loop enforces quality convergence:

  1. Implementing agent works on the task
  2. Review agents challenge in parallel
  3. If any [DEFECT] found → loop continues with fixes
  4. If no [DEFECT] remains → done
  5. Max iterations (default: 10) as safety cap

The implementing agent can't declare "done" alone — the adversarial review is the quality gate.

Agent memory

Each agent maintains persistent memory (.claude/agent-memory/<agent>/MEMORY.md) that calibrates over time:

  • Project-specific patterns and conventions
  • What challenges were accepted vs. overruled
  • Quality benchmarks learned from the codebase

Shared team learnings are stored in .claude/dev-team-learnings.md.

Customization

After installation, you can:

  • Edit agent prompts in .claude/agents/ to adjust focus areas or challenge style
  • Disable hooks by removing entries from .claude/settings.json
  • Add your own agents following the same frontmatter format
  • Tune agent memory in .claude/agent-memory/ to accelerate calibration

What gets installed

.claude/
  agents/           # Agent definitions (YAML frontmatter + prompt)
  hooks/            # Quality enforcement scripts
  skills/           # Skill definitions
  agent-memory/     # Per-agent persistent memory
  settings.json     # Hook configuration
CLAUDE.md           # Project instructions (dev-team section added via markers)

Contributing

  1. Every piece of work starts with a GitHub Issue
  2. Branch naming: feat/123-description or fix/456-description
  3. Commits reference issues: fixes #123 or refs #123
  4. All merges via PR — no direct pushes to main
  5. Run npm test before pushing

Development

npm install          # Install dependencies (dev only, zero runtime deps)
npm run build        # Compile TypeScript
npm test             # Build + run all tests
npm run lint         # Run oxlint
npm run format       # Run oxfmt

Architecture decisions are documented in docs/adr/.

License

MIT