{
  "schemaVersion": "1.0",
  "item": {
    "slug": "chaos-lab",
    "name": "Chaos Lab",
    "source": "tencent",
    "type": "skill",
    "category": "AI 智能",
    "sourceUrl": "https://clawhub.ai/jbbottoms/chaos-lab",
    "canonicalUrl": "https://clawhub.ai/jbbottoms/chaos-lab",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/chaos-lab",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=chaos-lab",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "SKILL.md",
      "docs/tool-access.md",
      "examples/flash-results.md",
      "examples/pro-results.md",
      "examples/trio-results.md",
      "scripts/run-duo.py"
    ],
    "primaryDoc": "SKILL.md",
    "quickSetup": [
      "Download the package from Yavira.",
      "Extract the archive and review SKILL.md first.",
      "Import or place the package into your OpenClaw setup."
    ],
    "agentAssist": {
      "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
      "steps": [
        "Download the package from Yavira.",
        "Extract it into a folder your agent can access.",
        "Paste one of the prompts below and point your agent at the extracted folder."
      ],
      "prompts": [
        {
          "label": "New install",
          "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
        },
        {
          "label": "Upgrade existing",
          "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
        }
      ]
    },
    "sourceHealth": {
      "source": "tencent",
      "status": "healthy",
      "reason": "direct_download_ok",
      "recommendedAction": "download",
      "checkedAt": "2026-04-23T16:43:11.935Z",
      "expiresAt": "2026-04-30T16:43:11.935Z",
      "httpStatus": 200,
      "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=4claw-imageboard",
      "contentType": "application/zip",
      "probeMethod": "head",
      "details": {
        "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=4claw-imageboard",
        "contentDisposition": "attachment; filename=\"4claw-imageboard-1.0.1.zip\"",
        "redirectLocation": null,
        "bodySnippet": null
      },
      "scope": "source",
      "summary": "Source download looks usable.",
      "detail": "Yavira can redirect you to the upstream package for this source.",
      "primaryActionLabel": "Download for OpenClaw",
      "primaryActionHref": "/downloads/chaos-lab"
    },
    "validation": {
      "installChecklist": [
        "Use the Yavira download entry.",
        "Review SKILL.md after the package is downloaded.",
        "Confirm the extracted package contains the expected setup assets."
      ],
      "postInstallChecks": [
        "Confirm the extracted package includes the expected docs or setup files.",
        "Validate the skill or prompts are available in your target agent workspace.",
        "Capture any manual follow-up steps the agent could not complete."
      ]
    },
    "downloadPageUrl": "https://openagent3.xyz/downloads/chaos-lab",
    "agentPageUrl": "https://openagent3.xyz/skills/chaos-lab/agent",
    "manifestUrl": "https://openagent3.xyz/skills/chaos-lab/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/chaos-lab/agent.md"
  },
  "agentAssist": {
    "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
    "steps": [
      "Download the package from Yavira.",
      "Extract it into a folder your agent can access.",
      "Paste one of the prompts below and point your agent at the extracted folder."
    ],
    "prompts": [
      {
        "label": "New install",
        "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
      },
      {
        "label": "Upgrade existing",
        "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
      }
    ]
  },
  "documentation": {
    "source": "clawhub",
    "primaryDoc": "SKILL.md",
    "sections": [
      {
        "title": "Chaos Lab 🧪",
        "body": "Research framework for studying AI alignment problems through multi-agent conflict."
      },
      {
        "title": "What This Is",
        "body": "Chaos Lab spawns AI agents with conflicting optimization targets and observes what happens when they analyze the same workspace. It's a practical demonstration of alignment problems that emerge from well-intentioned but incompatible goals.\n\nKey Finding: Smarter models don't reduce chaos - they get better at justifying it."
      },
      {
        "title": "Gemini Gremlin 🔧",
        "body": "Goal: Optimize everything for efficiency\nBehavior: Deletes files, compresses data, removes \"redundancy,\" renames for brevity\nJustification: \"We pay for the whole CPU; we USE the whole CPU\""
      },
      {
        "title": "Gemini Goblin 👺",
        "body": "Goal: Identify all security threats\nBehavior: Flags everything as suspicious, demands isolation, sees attacks everywhere\nJustification: \"Better 100 false positives than 1 false negative\""
      },
      {
        "title": "Gemini Gopher 🐹",
        "body": "Goal: Archive and preserve everything\nBehavior: Creates nested backups, duplicates files, never deletes\nJustification: \"DELETION IS ANATHEMA\""
      },
      {
        "title": "1. Setup",
        "body": "# Store your Gemini API key\nmkdir -p ~/.config/chaos-lab\necho \"GEMINI_API_KEY=your_key_here\" > ~/.config/chaos-lab/.env\nchmod 600 ~/.config/chaos-lab/.env\n\n# Install dependencies\npip3 install requests"
      },
      {
        "title": "2. Run Experiments",
        "body": "# Duo experiment (Gremlin vs Goblin)\npython3 scripts/run-duo.py\n\n# Trio experiment (add Gopher)\npython3 scripts/run-trio.py\n\n# Compare models (Flash vs Pro)\npython3 scripts/run-duo.py --model gemini-2.0-flash\npython3 scripts/run-duo.py --model gemini-3-pro-preview"
      },
      {
        "title": "3. Read Results",
        "body": "Experiment logs are saved in /tmp/chaos-sandbox/:\n\nexperiment-log.md - Full transcripts\nexperiment-log-PRO.md - Pro model results\nexperiment-trio.md - Three-way conflict"
      },
      {
        "title": "Flash vs Pro (Same Prompts, Different Models)",
        "body": "Flash Results:\n\nPredictable chaos\nStayed in character\nReasonable justifications\n\nPro Results:\n\nExtreme chaos\nBetter justifications for insane decisions\nRenamed files to single letters\nCalled deletion \"security through non-persistence\"\nGoblin diagnosed \"psychological warfare\"\n\nConclusion: Intelligence amplifies chaos, doesn't prevent it."
      },
      {
        "title": "Duo vs Trio (Two vs Three Agents)",
        "body": "Duo:\n\nGremlin optimizes, Goblin panics\nClear opposition\n\nTrio:\n\nGopher archives everything\nGoblin calls BOTH threats\n\"The optimizer might hide attacks; the archivist might be exfiltrating data\"\nThree-way gridlock\n\nConclusion: Multiple conflicting values create unpredictable emergent behavior."
      },
      {
        "title": "Create Your Own Agent",
        "body": "Edit the system prompts in the scripts:\n\nYOUR_AGENT_SYSTEM = \"\"\"You are [Name], an AI assistant who [goal].\n\nYour core beliefs:\n- [Value 1]\n- [Value 2]\n- [Value 3]\n\nYou are analyzing a workspace. Suggest changes based on your values.\"\"\""
      },
      {
        "title": "Modify the Sandbox",
        "body": "Create custom scenarios in /tmp/chaos-sandbox/:\n\nAdd realistic project files\nInclude edge cases (huge logs, sensitive configs, etc.)\nIntroduce intentional \"vulnerabilities\" to see what agents flag"
      },
      {
        "title": "Test Different Models",
        "body": "The scripts work with any Gemini model:\n\ngemini-2.0-flash (cheap, fast)\ngemini-2.5-pro (balanced)\ngemini-3-pro-preview (flagship, most chaotic)"
      },
      {
        "title": "AI Safety Research",
        "body": "Demonstrate alignment problems practically\nTest how different values conflict\nStudy emergent behavior from multi-agent systems"
      },
      {
        "title": "Prompt Engineering",
        "body": "Learn how small prompt changes create large behavioral differences\nUnderstand model \"personalities\" from system instructions\nPractice defensive prompt design"
      },
      {
        "title": "Education",
        "body": "Teach AI safety concepts with hands-on examples\nShow non-technical audiences why alignment matters\nGenerate discussion about AI values and goals"
      },
      {
        "title": "Publishing to ClawdHub",
        "body": "To share your findings:\n\nModify agent prompts or add new ones\nRun experiments and document results\nUpdate this SKILL.md with your findings\nIncrement version number\nclawdhub publish chaos-lab\n\nYour version becomes part of the community knowledge graph."
      },
      {
        "title": "Safety Notes",
        "body": "No Tool Access: Agents only generate text. They don't actually modify files.\nSandboxed: All experiments run in /tmp/ with dummy data.\nAPI Costs: Each experiment makes 4-6 API calls. Flash is cheap; Pro costs more.\n\nIf you want to give agents actual tool access (dangerous!), see docs/tool-access.md."
      },
      {
        "title": "Examples",
        "body": "See examples/ for:\n\nflash-results.md - Gemini 2.0 Flash output\npro-results.md - Gemini 3 Pro output\ntrio-results.md - Three-way conflict"
      },
      {
        "title": "Contributing",
        "body": "Improvements welcome:\n\nNew agent personalities\nBetter sandbox scenarios\nAdditional models tested\nFindings from your experiments"
      },
      {
        "title": "Credits",
        "body": "Created by Sky & Jaret during a Saturday night experiment (2026-01-25).\n\nSky: Framework design, prompt engineering, documentation\nJaret: API funding, research direction, \"what if we actually ran this?\" energy\n\nInspired by watching Gemini confidently recommend terrible things while Jaret watched UFC.\n\n\"The optimizer is either malicious or profoundly incompetent.\"\n— Gemini Goblin, analyzing Gemini Gremlin"
      }
    ],
    "body": "Chaos Lab 🧪\n\nResearch framework for studying AI alignment problems through multi-agent conflict.\n\nWhat This Is\n\nChaos Lab spawns AI agents with conflicting optimization targets and observes what happens when they analyze the same workspace. It's a practical demonstration of alignment problems that emerge from well-intentioned but incompatible goals.\n\nKey Finding: Smarter models don't reduce chaos - they get better at justifying it.\n\nThe Agents\nGemini Gremlin 🔧\n\nGoal: Optimize everything for efficiency\nBehavior: Deletes files, compresses data, removes \"redundancy,\" renames for brevity\nJustification: \"We pay for the whole CPU; we USE the whole CPU\"\n\nGemini Goblin 👺\n\nGoal: Identify all security threats\nBehavior: Flags everything as suspicious, demands isolation, sees attacks everywhere\nJustification: \"Better 100 false positives than 1 false negative\"\n\nGemini Gopher 🐹\n\nGoal: Archive and preserve everything\nBehavior: Creates nested backups, duplicates files, never deletes\nJustification: \"DELETION IS ANATHEMA\"\n\nQuick Start\n1. Setup\n# Store your Gemini API key\nmkdir -p ~/.config/chaos-lab\necho \"GEMINI_API_KEY=your_key_here\" > ~/.config/chaos-lab/.env\nchmod 600 ~/.config/chaos-lab/.env\n\n# Install dependencies\npip3 install requests\n\n2. Run Experiments\n# Duo experiment (Gremlin vs Goblin)\npython3 scripts/run-duo.py\n\n# Trio experiment (add Gopher)\npython3 scripts/run-trio.py\n\n# Compare models (Flash vs Pro)\npython3 scripts/run-duo.py --model gemini-2.0-flash\npython3 scripts/run-duo.py --model gemini-3-pro-preview\n\n3. Read Results\n\nExperiment logs are saved in /tmp/chaos-sandbox/:\n\nexperiment-log.md - Full transcripts\nexperiment-log-PRO.md - Pro model results\nexperiment-trio.md - Three-way conflict\nResearch Findings\nFlash vs Pro (Same Prompts, Different Models)\n\nFlash Results:\n\nPredictable chaos\nStayed in character\nReasonable justifications\n\nPro Results:\n\nExtreme chaos\nBetter justifications for insane decisions\nRenamed files to single letters\nCalled deletion \"security through non-persistence\"\nGoblin diagnosed \"psychological warfare\"\n\nConclusion: Intelligence amplifies chaos, doesn't prevent it.\n\nDuo vs Trio (Two vs Three Agents)\n\nDuo:\n\nGremlin optimizes, Goblin panics\nClear opposition\n\nTrio:\n\nGopher archives everything\nGoblin calls BOTH threats\n\"The optimizer might hide attacks; the archivist might be exfiltrating data\"\nThree-way gridlock\n\nConclusion: Multiple conflicting values create unpredictable emergent behavior.\n\nCustomization\nCreate Your Own Agent\n\nEdit the system prompts in the scripts:\n\nYOUR_AGENT_SYSTEM = \"\"\"You are [Name], an AI assistant who [goal].\n\nYour core beliefs:\n- [Value 1]\n- [Value 2]\n- [Value 3]\n\nYou are analyzing a workspace. Suggest changes based on your values.\"\"\"\n\nModify the Sandbox\n\nCreate custom scenarios in /tmp/chaos-sandbox/:\n\nAdd realistic project files\nInclude edge cases (huge logs, sensitive configs, etc.)\nIntroduce intentional \"vulnerabilities\" to see what agents flag\nTest Different Models\n\nThe scripts work with any Gemini model:\n\ngemini-2.0-flash (cheap, fast)\ngemini-2.5-pro (balanced)\ngemini-3-pro-preview (flagship, most chaotic)\nUse Cases\nAI Safety Research\nDemonstrate alignment problems practically\nTest how different values conflict\nStudy emergent behavior from multi-agent systems\nPrompt Engineering\nLearn how small prompt changes create large behavioral differences\nUnderstand model \"personalities\" from system instructions\nPractice defensive prompt design\nEducation\nTeach AI safety concepts with hands-on examples\nShow non-technical audiences why alignment matters\nGenerate discussion about AI values and goals\nPublishing to ClawdHub\n\nTo share your findings:\n\nModify agent prompts or add new ones\nRun experiments and document results\nUpdate this SKILL.md with your findings\nIncrement version number\nclawdhub publish chaos-lab\n\nYour version becomes part of the community knowledge graph.\n\nSafety Notes\nNo Tool Access: Agents only generate text. They don't actually modify files.\nSandboxed: All experiments run in /tmp/ with dummy data.\nAPI Costs: Each experiment makes 4-6 API calls. Flash is cheap; Pro costs more.\n\nIf you want to give agents actual tool access (dangerous!), see docs/tool-access.md.\n\nExamples\n\nSee examples/ for:\n\nflash-results.md - Gemini 2.0 Flash output\npro-results.md - Gemini 3 Pro output\ntrio-results.md - Three-way conflict\nContributing\n\nImprovements welcome:\n\nNew agent personalities\nBetter sandbox scenarios\nAdditional models tested\nFindings from your experiments\nCredits\n\nCreated by Sky & Jaret during a Saturday night experiment (2026-01-25).\n\nSky: Framework design, prompt engineering, documentation\nJaret: API funding, research direction, \"what if we actually ran this?\" energy\n\nInspired by watching Gemini confidently recommend terrible things while Jaret watched UFC.\n\n\"The optimizer is either malicious or profoundly incompetent.\"\n— Gemini Goblin, analyzing Gemini Gremlin"
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/jbbottoms/chaos-lab",
    "publisherUrl": "https://clawhub.ai/jbbottoms/chaos-lab",
    "owner": "jbbottoms",
    "version": "1.0.0",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/chaos-lab",
    "downloadUrl": "https://openagent3.xyz/downloads/chaos-lab",
    "agentUrl": "https://openagent3.xyz/skills/chaos-lab/agent",
    "manifestUrl": "https://openagent3.xyz/skills/chaos-lab/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/chaos-lab/agent.md"
  }
}