{
  "markdown": "# Challenger Plugin\n\nA Claude Code plugin that stress-tests your code, reasoning, and decisions through multi-agent adversarial challenge rounds. Instead of accepting conclusions at face value, Challenger assembles a panel of specialized adversaries that poke holes, gather evidence, and force refinement until your resolution is bulletproof.\n\n## How It Works\n\nWhen you invoke `/challenge`, the plugin:\n\n1. **Classifies** what you're challenging (security, architecture, cost/complexity, or general reasoning)\n2. **Auto-scales** intensity based on stakes — quick sanity check, deep review, or brutal stress test\n3. **Dispatches parallel agents** — real subagents running simultaneously, each with isolated context and independent perspectives\n4. **Runs multi-round refinement** silently — agents challenge, evidence is gathered, resolution is refined\n5. **Delivers one concise result** — the battle-tested resolution with key challenges, evidence, and confidence score\n\n## Agents\n\n| Agent | Lens | What They Challenge |\n|-------|------|-------------------|\n| **Skeptic** | Logic & reasoning | Weak assumptions, logical fallacies, unverified claims |\n| **Sentinel** | Security & risk | Threat models, attack vectors, data exposure, auth gaps |\n| **Architect** | Design & scalability | Over-engineering, coupling, missing abstractions, scaling limits |\n| **Pragmatist** | Cost & complexity | YAGNI violations, premature optimization, simpler alternatives |\n| **Challenger Watch** | Advisory | Passively suggests `/challenge` during high-stakes moments |\n\nEvery challenge is **evidence-backed** — agents read your code, search for patterns, check git history, and run existing tests to support or dismiss their challenges. No hand-waving.\n\n## Skills\n\n### `/challenge`\n\nThe main skill. Runs adversarial challenge rounds against a resolution.\n\n```\n/challenge                              # Challenge the last conclusion in conversation\n/challenge <topic>                      # Challenge a specific topic\n/challenge --agents skeptic,sentinel    # Pick which agents to use\n/challenge --intensity brutal           # Override auto-scaling\n```\n\n**Intensity levels:**\n\n| Level | Rounds | Agents | When |\n|-------|--------|--------|------|\n| Quick | 1 | 1-2 | Minor decisions, simple claims |\n| Deep | up to 3 | 2-3 | Architecture decisions, security-sensitive code |\n| Brutal | no cap | All 4 | Production deploys, breaking changes, critical paths |\n\n### `/challenge-report`\n\nSaves the results of the last challenge session to a markdown file at `docs/challenge-reports/`.\n\n```\n/challenge-report\n```\n\nGenerates a report with: final resolution, per-agent scores, round summaries, dismissed challenges, and evidence trail.\n\n## Installation\n\n### From GitHub\n\n```bash\n# Inside a Claude Code session:\n/plugin marketplace add ismailnakkar/challenger-plugin\n/plugin install challenger-plugin@challenger\n```\n\n### Local Development\n\n```bash\nclaude --plugin-dir /path/to/challenger-plugin\n```\n\n## Example\n\n```\nYou: /challenge Our API should use JWT tokens for authentication\n\nChallenge initiated\nTopic: JWT token authentication for API\nCategory: Security\nIntensity: Deep — security-sensitive authentication decision\nAgents: skeptic, sentinel, architect\n\n[agents dispatched in parallel, running silently...]\n\n## ✅ Challenge Complete\n\nResolution: Use JWT tokens with short-lived access tokens (15min),\nHTTP-only refresh tokens, and server-side revocation via a token\nblocklist. Store secrets in environment variables, not code.\n\nConfidence: 8/10\nRounds: 2 | Intensity: Deep | Agents: skeptic, sentinel, architect\n\nKey challenges that shaped this resolution:\n- 🔴 sentinel: JWT in localStorage is XSS-vulnerable → moved to HTTP-only cookies\n- 🟡 skeptic: No revocation strategy for compromised tokens → added token blocklist\n- 🔵 architect: Refresh token rotation missing → added rotation on each refresh\n\nSurviving open questions:\n- Token blocklist storage strategy (Redis vs DB) not yet decided\n\nEvidence:\n- src/auth/middleware.ts:42 — current token validation has no expiry check\n```\n\n## Project Structure\n\n```\nchallenger-plugin/\n├── .claude-plugin/\n│   ├── plugin.json           # Plugin manifest\n│   └── marketplace.json      # Marketplace config\n├── agents/\n│   ├── skeptic.md            # Logic & reasoning challenger\n│   ├── sentinel.md           # Security & risk challenger\n│   ├── architect.md          # Design & scalability challenger\n│   ├── pragmatist.md         # Cost & complexity challenger\n│   └── challenger-watch.md   # Passive auto-suggest advisor\n└── skills/\n    ├── challenge/\n    │   └── SKILL.md          # Multi-agent orchestrator\n    └── challenge-report/\n        └── SKILL.md          # Report generator\n```\n\n## License\n\nMIT\n",
  "bytes": 4731,
  "sha": "9cf37e05511da0b58c0a673012ca75a397cab7885277c5ffc8d7d15e5b29f608",
  "repo_slug": "ismailnakkar/challenger-plugin",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_ismailnakkar_challenger_plugin_challenge_0d03dde1/readme"
}