{
  "schemaVersion": "1.0",
  "item": {
    "slug": "multi-agent-orchestration",
    "name": "Colony",
    "source": "tencent",
    "type": "skill",
    "category": "AI 智能",
    "sourceUrl": "https://clawhub.ai/dabruhce/multi-agent-orchestration",
    "canonicalUrl": "https://clawhub.ai/dabruhce/multi-agent-orchestration",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/multi-agent-orchestration",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=multi-agent-orchestration",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "COLONY-DOCS.md",
      "package.json",
      "SKILL.md",
      "scripts/agent-wrapper.mjs",
      "scripts/audit.mjs",
      "scripts/colony-worker.mjs"
    ],
    "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": "2026-04-30T16:55:25.780Z",
      "expiresAt": "2026-05-07T16:55:25.780Z",
      "httpStatus": 200,
      "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
      "contentType": "application/zip",
      "probeMethod": "head",
      "details": {
        "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
        "contentDisposition": "attachment; filename=\"network-1.0.0.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/multi-agent-orchestration"
    },
    "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/multi-agent-orchestration",
    "agentPageUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/agent",
    "manifestUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/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": "Colony Orchestration Skill",
        "body": "Multi-agent task delegation and process orchestration with audit logging and agent learning capabilities."
      },
      {
        "title": "Quick Start",
        "body": "# Single task - auto-routed\nnode scripts/colony.mjs dispatch \"find top 5 time-series databases\"\n\n# Multi-stage process\nnode scripts/colony.mjs process validate-idea --context \"AI meal planning for parents\"\nnode scripts/colony.mjs process-status  # check progress\nnode scripts/colony.mjs approve abc123  # continue past checkpoint\n\n# Check audit stats\nnode scripts/colony.mjs audit\n\n# View agent memory\nnode scripts/colony.mjs memory scout"
      },
      {
        "title": "Agents",
        "body": "AgentRoleSpecializationscuttleresearcherQuick searches, lookups, fact-findingscoutresearcherDeep market/competitor research, intelligenceforecastanalystData analysis, trends, projectionspincercoderWriting, debugging, refactoring codeshellopsGit, deployments, system tasksforgeproductPRDs, specs, roadmapsledgerfinancePricing, costs, business casesmusecreativeBrainstorming, naming, ideasscribewriterBlog posts, docs, long-form contentquillcopywriterLanding pages, sales copy, adsechosocialTweets, social posts, promotionsentryqaTesting, bug verification"
      },
      {
        "title": "Dispatch (Auto-Route)",
        "body": "node scripts/colony.mjs dispatch \"research best practices for API rate limiting\"\n\nAutomatically detects the best agent based on task keywords."
      },
      {
        "title": "Assign to Specific Agent",
        "body": "node scripts/colony.mjs assign scout \"find top 5 time-series databases\"\nnode scripts/colony.mjs assign pincer \"refactor the auth module to use JWT\"\nnode scripts/colony.mjs assign shell \"deploy the staging branch\""
      },
      {
        "title": "Check Status",
        "body": "node scripts/colony.mjs status\n\nShows all agents and their current tasks."
      },
      {
        "title": "Get Results",
        "body": "node scripts/colony.mjs results              # Latest completed task\nnode scripts/colony.mjs results abc123       # Specific task by ID"
      },
      {
        "title": "View History",
        "body": "node scripts/colony.mjs history              # Last 10 completed/failed\nnode scripts/colony.mjs history --limit 20   # Custom limit"
      },
      {
        "title": "Process Commands",
        "body": "Processes are multi-stage workflows that chain agents together."
      },
      {
        "title": "List Available Processes",
        "body": "node scripts/colony.mjs processes"
      },
      {
        "title": "Start a Process",
        "body": "node scripts/colony.mjs process <process-name> --context \"description\"\n\nExamples:\n\nnode scripts/colony.mjs process validate-idea --context \"AI-powered meal planning for busy parents\"\nnode scripts/colony.mjs process content-pipeline --context \"How to use vector databases for RAG\"\nnode scripts/colony.mjs process product-launch --context \"Life Lunch ritual kit for parents\"\nnode scripts/colony.mjs process bug-triage --context \"Login fails with OAuth on mobile\""
      },
      {
        "title": "Check Process Status",
        "body": "node scripts/colony.mjs process-status           # Show latest run\nnode scripts/colony.mjs process-status abc123    # Specific run\n\nShows: current stage, completed stages, checkpoints, output files."
      },
      {
        "title": "View Process Runs",
        "body": "node scripts/colony.mjs runs                 # All runs (active, paused, completed)\nnode scripts/colony.mjs runs --limit 5       # Last 5"
      },
      {
        "title": "Approve Checkpoint",
        "body": "When a process hits a checkpoint, it pauses for human approval:\n\nnode scripts/colony.mjs approve abc123\n\nAlso used to retry a failed stage."
      },
      {
        "title": "Cancel a Process",
        "body": "node scripts/colony.mjs cancel abc123"
      },
      {
        "title": "Audit Commands",
        "body": "Track agent performance, task statistics, and system health."
      },
      {
        "title": "Dashboard",
        "body": "node scripts/colony.mjs audit\n\nShows global stats, per-agent summary, and recent events."
      },
      {
        "title": "Agent Details",
        "body": "node scripts/colony.mjs audit agent scout\nnode scripts/colony.mjs audit agent pincer\n\nShows detailed stats for a specific agent including:\n\nTotal tasks, success rate\nAverage duration\nToken usage\nRecent failures"
      },
      {
        "title": "Event Log",
        "body": "node scripts/colony.mjs audit log              # Last 20 events\nnode scripts/colony.mjs audit log --limit 50   # More events"
      },
      {
        "title": "Slowest Tasks",
        "body": "node scripts/colony.mjs audit slow             # Top 10 slowest\nnode scripts/colony.mjs audit slow --limit 20"
      },
      {
        "title": "Recent Failures",
        "body": "node scripts/colony.mjs audit failures         # Last 10 failures\nnode scripts/colony.mjs audit failures --limit 20"
      },
      {
        "title": "Learning Commands",
        "body": "Agents learn from experience and share knowledge."
      },
      {
        "title": "Feedback",
        "body": "Record feedback for completed tasks:\n\nnode scripts/colony.mjs feedback abc123 \"Great research, but needed more pricing data\""
      },
      {
        "title": "Agent Memory",
        "body": "Each agent has a persistent memory file with lessons learned:\n\n# View an agent's memory\nnode scripts/colony.mjs memory scout\n\n# Add a lesson\nnode scripts/colony.mjs memory scout add \"Always check publication dates on research sources\"\n\n# Add to specific sections\nnode scripts/colony.mjs memory scout add \"Use bullet points for clarity\" --pattern\nnode scripts/colony.mjs memory scout add \"Missed competitor X in analysis\" --mistake\nnode scripts/colony.mjs memory scout add \"Prefers markdown tables over lists\" --pref"
      },
      {
        "title": "Shared Learnings",
        "body": "Cross-agent insights and lessons:\n\n# View all shared learnings\nnode scripts/colony.mjs learn\n\n# Add a learning\nnode scripts/colony.mjs learn add \"validate-idea works better with 3 competitors max\" --category process\nnode scripts/colony.mjs learn add \"Always verify API rate limits early\" --category technical --source run-abc123"
      },
      {
        "title": "Global Context",
        "body": "Shared context all agents can access:\n\n# View global context\nnode scripts/colony.mjs context\n\n# Set preferences\nnode scripts/colony.mjs context set preferences.codeStyle \"TypeScript, functional\"\nnode scripts/colony.mjs context set preferences.timezone \"America/Chicago\"\n\n# Add active facts (temporary context)\nnode scripts/colony.mjs context add-fact \"We're targeting enterprise customers\"\nnode scripts/colony.mjs context add-fact \"Launch deadline is Q2 2024\"\n\n# Add decisions\nnode scripts/colony.mjs context add-decision \"Use Postgres over MySQL\" --project \"life-lunch\"\n\n# Add projects\nnode scripts/colony.mjs context add-project \"life-lunch\""
      },
      {
        "title": "Retrospective",
        "body": "Review recent activity and generate insights:\n\nnode scripts/colony.mjs retro              # Last 7 days\nnode scripts/colony.mjs retro --days 14    # Last 14 days\n\nShows:\n\nTask completion summary\nPer-agent stats\nFailure patterns\nSuggested learnings"
      },
      {
        "title": "validate-idea",
        "body": "Validate a business idea end-to-end\n\nStages: brainstorm → research → analyze → spec → estimate\nCheckpoints: after analyze\nOutput: business-case.md"
      },
      {
        "title": "product-launch",
        "body": "End-to-end product launch\n\nStages: research → spec → build → copy\nCheckpoints: after spec, after copy\nOutput: market-brief.md, prd.md, code/, landing-copy.md"
      },
      {
        "title": "content-pipeline",
        "body": "Research, write, publish, promote content\n\nStages: research → draft → review → publish → promote\nCheckpoints: review (human reviews draft)\nOutput: research.md, draft.md, social-posts.md"
      },
      {
        "title": "bug-triage",
        "body": "Reproduce, fix, deploy bug fixes\n\nStages: reproduce → fix → test → deploy\nCheckpoints: none (fast path)\nOutput: bug-report.md, fix-summary.md"
      },
      {
        "title": "customer-research",
        "body": "Deep dive on a customer segment\n\nStages: identify → pain-points → validate → synthesize\nCheckpoints: none\nOutput: customer-profile.md, insights.md"
      },
      {
        "title": "landing-page",
        "body": "Create a full landing page\n\nStages: strategy → copy → review → build\nCheckpoints: after copy review\nOutput: strategy.md, copy.md, landing.html, landing.css"
      },
      {
        "title": "How Processes Work",
        "body": "Start - Process creates a run entry and spawns first stage agent\nExecute - Each stage runs with inputs from previous stages\nCheckpoint - If stage is a checkpoint, process pauses for approval\nContinue - After approval, next stage runs\nComplete - All stages done, outputs in colony/context/<run-id>/"
      },
      {
        "title": "Context Passing",
        "body": "{context} in task templates is replaced with your --context value\nStage outputs are saved to colony/context/<run-id>/<output-file>\nNext stage reads inputs from previous stage's output files\nAgent memory and global context are injected into prompts\nFull task history in tasks.json"
      },
      {
        "title": "Parallel Stages",
        "body": "Stages that share the same parallel_group run concurrently:\n\nstages:\n  - id: spec\n    agent: forge\n    inputs: [analysis.md]\n    parallel_group: \"final\"  # Stages with same group run together\n    \n  - id: estimate\n    agent: ledger\n    inputs: [analysis.md]\n    parallel_group: \"final\"  # Same group = parallel execution\n\nWhen the process reaches a parallel group:\n\nAll consecutive stages with the same parallel_group are collected\nAll stages spawn concurrently (using Promise.all())\nProcess waits for ALL parallel stages to complete\nIf any stage fails, the entire group fails\nCheckpoints work per-group (pause after all parallel stages complete)\n\nOutput shows parallel execution clearly:\n\n═══ Parallel Group: final (2 stages) ═══\n    → Stage 4: spec (forge)\n    → Stage 5: estimate (ledger)\n\n--- [PARALLEL] Stage 4/5: spec ---\n--- [PARALLEL] Stage 5/5: estimate ---\n\n═══ Parallel Group: final completed ═══\n\nWhen to use parallel groups:\n\nStages that read the same inputs (no dependencies on each other)\nBuild + copy tasks (both depend on spec, not on each other)\nMultiple analyses of the same data\nIndependent research threads\n\nProcesses with parallel stages:\n\nvalidate-idea: spec + estimate run in parallel\nproduct-launch: build + copy run in parallel"
      },
      {
        "title": "Notifications",
        "body": "Colony can send notifications when processes hit checkpoints, complete, or fail. Notifications use openclaw cron wake to alert you.\n\nConfiguration (colony/config.yaml):\n\nnotifications:\n  enabled: true         # Master switch for all notifications\n  on_checkpoint: true   # Notify when process pauses at checkpoint\n  on_complete: true     # Notify when process finishes\n  on_failure: true      # Notify when process/stage fails\n\nManage via CLI:\n\n# View current config\nnode scripts/colony.mjs config\n\n# Disable all notifications\nnode scripts/colony.mjs config set notifications.enabled false\n\n# Enable only failure notifications\nnode scripts/colony.mjs config set notifications.on_checkpoint false\nnode scripts/colony.mjs config set notifications.on_complete false\nnode scripts/colony.mjs config set notifications.on_failure true\n\nNotification examples:\n\n🛑 Colony checkpoint: Process \"validate-idea\" paused after stage \"analyze\". To continue: colony approve abc123\n✅ Colony complete: Process \"validate-idea\" finished in 120s. Run ID: abc123\n❌ Colony failed: Process \"validate-idea\" failed at stage \"research\". Error: Agent timed out. Run ID: abc123"
      },
      {
        "title": "Checkpoints",
        "body": "Checkpoints pause the process for human review. Two ways to define:\n\nIn process checkpoints array (after that stage completes)\nAs a standalone stage with checkpoint: true (human-only review step)"
      },
      {
        "title": "File Structure",
        "body": "skills/colony/\n├── SKILL.md              # This file\n├── package.json          # Dependencies (js-yaml)\n├── colony/\n│   ├── agents.yaml       # Agent definitions\n│   ├── processes.yaml    # Process definitions\n│   ├── config.yaml       # Notification & behavior config\n│   ├── tasks.json        # Task queue and history\n│   ├── runs.json         # Process run tracking\n│   ├── feedback.json     # Task feedback storage\n│   ├── learnings.yaml    # Shared cross-agent learnings\n│   ├── global-context.json  # Shared context for all agents\n│   ├── audit/\n│   │   ├── log.jsonl     # Append-only event log\n│   │   ├── global.json   # Aggregate statistics\n│   │   └── agents/       # Per-agent statistics\n│   │       ├── scout.json\n│   │       ├── pincer.json\n│   │       └── ...\n│   ├── memory/           # Per-agent persistent memory\n│   │   ├── scout.md\n│   │   ├── pincer.md\n│   │   └── ...\n│   └── context/          # Per-task and per-run outputs\n│       └── <run-id>/\n└── scripts/\n    ├── colony.mjs         # Main CLI\n    ├── colony-worker.mjs  # Background agent executor\n    ├── agent-wrapper.mjs # Task lifecycle utilities\n    ├── audit.mjs         # Audit system functions\n    └── learning.mjs      # Learning system functions"
      },
      {
        "title": "Audit Events",
        "body": "The audit log tracks these events:\n\nEventFieldstask_startedtaskId, agent, processRunId?, stage?task_completedtaskId, agent, durationMs, tokens, successtask_failedtaskId, agent, durationMs, errorcheckpoint_waitingrunId, stagecheckpoint_approvedrunId, stagecheckpoint_rejectedrunId, stage, reasonprocess_startedrunId, processId, contextprocess_completedrunId, processId, durationMsfeedback_receivedtaskId, agent, feedback"
      },
      {
        "title": "Add New Agents",
        "body": "Edit colony/agents.yaml:\n\nagents:\n  myagent:\n    role: specialist\n    description: >\n      What this agent does...\n    model: anthropic/claude-sonnet-4\n    triggers:\n      - keyword1\n      - keyword2\n\nAfter adding, create their memory file:\n\ntouch colony/memory/myagent.md"
      },
      {
        "title": "Add New Processes",
        "body": "Edit colony/processes.yaml:\n\nprocesses:\n  my-process:\n    description: \"What this process does\"\n    triggers: [keyword1, keyword2]\n    stages:\n      - id: stage1\n        agent: scout\n        task: \"Do something with: {context}\"\n        outputs: [output1.md]\n      - id: stage2\n        agent: pincer\n        task: \"Next step based on previous\"\n        inputs: [output1.md]\n        outputs: [output2.md]\n    checkpoints: [stage1]  # Optional: pause after these stages"
      },
      {
        "title": "Integration",
        "body": "Works with OpenClaw's agent sessions.\n\nDispatch/Assign (async): Tasks are spawned in the background and the CLI returns immediately. Use colony status to monitor progress and colony results <task-id> to view output.\n\nProcess stages (blocking): Multi-stage processes run sequentially, waiting for each stage to complete before proceeding. This ensures proper data flow between stages and checkpoint handling.\n\nEach agent receives:\n\nTheir role description\nLessons from their memory file\nActive facts from global context\nProject/preference context"
      },
      {
        "title": "Validate a Startup Idea",
        "body": "node scripts/colony.mjs process validate-idea \\\n  --context \"Subscription box for home coffee brewing experiments\"\n\nWatch as it flows: brainstorm → research → analyze → (checkpoint) → spec → estimate"
      },
      {
        "title": "Write and Publish a Blog Post",
        "body": "node scripts/colony.mjs process content-pipeline \\\n  --context \"Why RAG is eating traditional search\"\n\nStages: research → draft → (human review) → publish → promote"
      },
      {
        "title": "Quick Research Task",
        "body": "node scripts/colony.mjs dispatch \"compare Pinecone vs Weaviate vs Milvus\"\n\nAuto-routes to scout, returns comparison."
      },
      {
        "title": "Track Performance",
        "body": "# After several tasks, check overall health\nnode scripts/colony.mjs audit\n\n# Deep dive into a struggling agent\nnode scripts/colony.mjs audit agent pincer\nnode scripts/colony.mjs audit failures\n\n# Add learnings from issues\nnode scripts/colony.mjs memory pincer add \"Handle file not found errors gracefully\" --mistake"
      }
    ],
    "body": "Colony Orchestration Skill\n\nMulti-agent task delegation and process orchestration with audit logging and agent learning capabilities.\n\nQuick Start\n# Single task - auto-routed\nnode scripts/colony.mjs dispatch \"find top 5 time-series databases\"\n\n# Multi-stage process\nnode scripts/colony.mjs process validate-idea --context \"AI meal planning for parents\"\nnode scripts/colony.mjs process-status  # check progress\nnode scripts/colony.mjs approve abc123  # continue past checkpoint\n\n# Check audit stats\nnode scripts/colony.mjs audit\n\n# View agent memory\nnode scripts/colony.mjs memory scout\n\nAgents\nAgent\tRole\tSpecialization\nscuttle\tresearcher\tQuick searches, lookups, fact-finding\nscout\tresearcher\tDeep market/competitor research, intelligence\nforecast\tanalyst\tData analysis, trends, projections\npincer\tcoder\tWriting, debugging, refactoring code\nshell\tops\tGit, deployments, system tasks\nforge\tproduct\tPRDs, specs, roadmaps\nledger\tfinance\tPricing, costs, business cases\nmuse\tcreative\tBrainstorming, naming, ideas\nscribe\twriter\tBlog posts, docs, long-form content\nquill\tcopywriter\tLanding pages, sales copy, ads\necho\tsocial\tTweets, social posts, promotion\nsentry\tqa\tTesting, bug verification\nTask Commands\nDispatch (Auto-Route)\nnode scripts/colony.mjs dispatch \"research best practices for API rate limiting\"\n\n\nAutomatically detects the best agent based on task keywords.\n\nAssign to Specific Agent\nnode scripts/colony.mjs assign scout \"find top 5 time-series databases\"\nnode scripts/colony.mjs assign pincer \"refactor the auth module to use JWT\"\nnode scripts/colony.mjs assign shell \"deploy the staging branch\"\n\nCheck Status\nnode scripts/colony.mjs status\n\n\nShows all agents and their current tasks.\n\nGet Results\nnode scripts/colony.mjs results              # Latest completed task\nnode scripts/colony.mjs results abc123       # Specific task by ID\n\nView History\nnode scripts/colony.mjs history              # Last 10 completed/failed\nnode scripts/colony.mjs history --limit 20   # Custom limit\n\nProcess Commands\n\nProcesses are multi-stage workflows that chain agents together.\n\nList Available Processes\nnode scripts/colony.mjs processes\n\nStart a Process\nnode scripts/colony.mjs process <process-name> --context \"description\"\n\n\nExamples:\n\nnode scripts/colony.mjs process validate-idea --context \"AI-powered meal planning for busy parents\"\nnode scripts/colony.mjs process content-pipeline --context \"How to use vector databases for RAG\"\nnode scripts/colony.mjs process product-launch --context \"Life Lunch ritual kit for parents\"\nnode scripts/colony.mjs process bug-triage --context \"Login fails with OAuth on mobile\"\n\nCheck Process Status\nnode scripts/colony.mjs process-status           # Show latest run\nnode scripts/colony.mjs process-status abc123    # Specific run\n\n\nShows: current stage, completed stages, checkpoints, output files.\n\nView Process Runs\nnode scripts/colony.mjs runs                 # All runs (active, paused, completed)\nnode scripts/colony.mjs runs --limit 5       # Last 5\n\nApprove Checkpoint\n\nWhen a process hits a checkpoint, it pauses for human approval:\n\nnode scripts/colony.mjs approve abc123\n\n\nAlso used to retry a failed stage.\n\nCancel a Process\nnode scripts/colony.mjs cancel abc123\n\nAudit Commands\n\nTrack agent performance, task statistics, and system health.\n\nDashboard\nnode scripts/colony.mjs audit\n\n\nShows global stats, per-agent summary, and recent events.\n\nAgent Details\nnode scripts/colony.mjs audit agent scout\nnode scripts/colony.mjs audit agent pincer\n\n\nShows detailed stats for a specific agent including:\n\nTotal tasks, success rate\nAverage duration\nToken usage\nRecent failures\nEvent Log\nnode scripts/colony.mjs audit log              # Last 20 events\nnode scripts/colony.mjs audit log --limit 50   # More events\n\nSlowest Tasks\nnode scripts/colony.mjs audit slow             # Top 10 slowest\nnode scripts/colony.mjs audit slow --limit 20\n\nRecent Failures\nnode scripts/colony.mjs audit failures         # Last 10 failures\nnode scripts/colony.mjs audit failures --limit 20\n\nLearning Commands\n\nAgents learn from experience and share knowledge.\n\nFeedback\n\nRecord feedback for completed tasks:\n\nnode scripts/colony.mjs feedback abc123 \"Great research, but needed more pricing data\"\n\nAgent Memory\n\nEach agent has a persistent memory file with lessons learned:\n\n# View an agent's memory\nnode scripts/colony.mjs memory scout\n\n# Add a lesson\nnode scripts/colony.mjs memory scout add \"Always check publication dates on research sources\"\n\n# Add to specific sections\nnode scripts/colony.mjs memory scout add \"Use bullet points for clarity\" --pattern\nnode scripts/colony.mjs memory scout add \"Missed competitor X in analysis\" --mistake\nnode scripts/colony.mjs memory scout add \"Prefers markdown tables over lists\" --pref\n\nShared Learnings\n\nCross-agent insights and lessons:\n\n# View all shared learnings\nnode scripts/colony.mjs learn\n\n# Add a learning\nnode scripts/colony.mjs learn add \"validate-idea works better with 3 competitors max\" --category process\nnode scripts/colony.mjs learn add \"Always verify API rate limits early\" --category technical --source run-abc123\n\nGlobal Context\n\nShared context all agents can access:\n\n# View global context\nnode scripts/colony.mjs context\n\n# Set preferences\nnode scripts/colony.mjs context set preferences.codeStyle \"TypeScript, functional\"\nnode scripts/colony.mjs context set preferences.timezone \"America/Chicago\"\n\n# Add active facts (temporary context)\nnode scripts/colony.mjs context add-fact \"We're targeting enterprise customers\"\nnode scripts/colony.mjs context add-fact \"Launch deadline is Q2 2024\"\n\n# Add decisions\nnode scripts/colony.mjs context add-decision \"Use Postgres over MySQL\" --project \"life-lunch\"\n\n# Add projects\nnode scripts/colony.mjs context add-project \"life-lunch\"\n\nRetrospective\n\nReview recent activity and generate insights:\n\nnode scripts/colony.mjs retro              # Last 7 days\nnode scripts/colony.mjs retro --days 14    # Last 14 days\n\n\nShows:\n\nTask completion summary\nPer-agent stats\nFailure patterns\nSuggested learnings\nAvailable Processes\nvalidate-idea\n\nValidate a business idea end-to-end\n\nStages: brainstorm → research → analyze → spec → estimate\nCheckpoints: after analyze\nOutput: business-case.md\nproduct-launch\n\nEnd-to-end product launch\n\nStages: research → spec → build → copy\nCheckpoints: after spec, after copy\nOutput: market-brief.md, prd.md, code/, landing-copy.md\ncontent-pipeline\n\nResearch, write, publish, promote content\n\nStages: research → draft → review → publish → promote\nCheckpoints: review (human reviews draft)\nOutput: research.md, draft.md, social-posts.md\nbug-triage\n\nReproduce, fix, deploy bug fixes\n\nStages: reproduce → fix → test → deploy\nCheckpoints: none (fast path)\nOutput: bug-report.md, fix-summary.md\ncustomer-research\n\nDeep dive on a customer segment\n\nStages: identify → pain-points → validate → synthesize\nCheckpoints: none\nOutput: customer-profile.md, insights.md\nlanding-page\n\nCreate a full landing page\n\nStages: strategy → copy → review → build\nCheckpoints: after copy review\nOutput: strategy.md, copy.md, landing.html, landing.css\nHow Processes Work\nStart - Process creates a run entry and spawns first stage agent\nExecute - Each stage runs with inputs from previous stages\nCheckpoint - If stage is a checkpoint, process pauses for approval\nContinue - After approval, next stage runs\nComplete - All stages done, outputs in colony/context/<run-id>/\nContext Passing\n{context} in task templates is replaced with your --context value\nStage outputs are saved to colony/context/<run-id>/<output-file>\nNext stage reads inputs from previous stage's output files\nAgent memory and global context are injected into prompts\nFull task history in tasks.json\nParallel Stages\n\nStages that share the same parallel_group run concurrently:\n\nstages:\n  - id: spec\n    agent: forge\n    inputs: [analysis.md]\n    parallel_group: \"final\"  # Stages with same group run together\n    \n  - id: estimate\n    agent: ledger\n    inputs: [analysis.md]\n    parallel_group: \"final\"  # Same group = parallel execution\n\n\nWhen the process reaches a parallel group:\n\nAll consecutive stages with the same parallel_group are collected\nAll stages spawn concurrently (using Promise.all())\nProcess waits for ALL parallel stages to complete\nIf any stage fails, the entire group fails\nCheckpoints work per-group (pause after all parallel stages complete)\n\nOutput shows parallel execution clearly:\n\n═══ Parallel Group: final (2 stages) ═══\n    → Stage 4: spec (forge)\n    → Stage 5: estimate (ledger)\n\n--- [PARALLEL] Stage 4/5: spec ---\n--- [PARALLEL] Stage 5/5: estimate ---\n\n═══ Parallel Group: final completed ═══\n\n\nWhen to use parallel groups:\n\nStages that read the same inputs (no dependencies on each other)\nBuild + copy tasks (both depend on spec, not on each other)\nMultiple analyses of the same data\nIndependent research threads\n\nProcesses with parallel stages:\n\nvalidate-idea: spec + estimate run in parallel\nproduct-launch: build + copy run in parallel\nNotifications\n\nColony can send notifications when processes hit checkpoints, complete, or fail. Notifications use openclaw cron wake to alert you.\n\nConfiguration (colony/config.yaml):\n\nnotifications:\n  enabled: true         # Master switch for all notifications\n  on_checkpoint: true   # Notify when process pauses at checkpoint\n  on_complete: true     # Notify when process finishes\n  on_failure: true      # Notify when process/stage fails\n\n\nManage via CLI:\n\n# View current config\nnode scripts/colony.mjs config\n\n# Disable all notifications\nnode scripts/colony.mjs config set notifications.enabled false\n\n# Enable only failure notifications\nnode scripts/colony.mjs config set notifications.on_checkpoint false\nnode scripts/colony.mjs config set notifications.on_complete false\nnode scripts/colony.mjs config set notifications.on_failure true\n\n\nNotification examples:\n\n🛑 Colony checkpoint: Process \"validate-idea\" paused after stage \"analyze\". To continue: colony approve abc123\n✅ Colony complete: Process \"validate-idea\" finished in 120s. Run ID: abc123\n❌ Colony failed: Process \"validate-idea\" failed at stage \"research\". Error: Agent timed out. Run ID: abc123\nCheckpoints\n\nCheckpoints pause the process for human review. Two ways to define:\n\nIn process checkpoints array (after that stage completes)\nAs a standalone stage with checkpoint: true (human-only review step)\nFile Structure\nskills/colony/\n├── SKILL.md              # This file\n├── package.json          # Dependencies (js-yaml)\n├── colony/\n│   ├── agents.yaml       # Agent definitions\n│   ├── processes.yaml    # Process definitions\n│   ├── config.yaml       # Notification & behavior config\n│   ├── tasks.json        # Task queue and history\n│   ├── runs.json         # Process run tracking\n│   ├── feedback.json     # Task feedback storage\n│   ├── learnings.yaml    # Shared cross-agent learnings\n│   ├── global-context.json  # Shared context for all agents\n│   ├── audit/\n│   │   ├── log.jsonl     # Append-only event log\n│   │   ├── global.json   # Aggregate statistics\n│   │   └── agents/       # Per-agent statistics\n│   │       ├── scout.json\n│   │       ├── pincer.json\n│   │       └── ...\n│   ├── memory/           # Per-agent persistent memory\n│   │   ├── scout.md\n│   │   ├── pincer.md\n│   │   └── ...\n│   └── context/          # Per-task and per-run outputs\n│       └── <run-id>/\n└── scripts/\n    ├── colony.mjs         # Main CLI\n    ├── colony-worker.mjs  # Background agent executor\n    ├── agent-wrapper.mjs # Task lifecycle utilities\n    ├── audit.mjs         # Audit system functions\n    └── learning.mjs      # Learning system functions\n\nAudit Events\n\nThe audit log tracks these events:\n\nEvent\tFields\ntask_started\ttaskId, agent, processRunId?, stage?\ntask_completed\ttaskId, agent, durationMs, tokens, success\ntask_failed\ttaskId, agent, durationMs, error\ncheckpoint_waiting\trunId, stage\ncheckpoint_approved\trunId, stage\ncheckpoint_rejected\trunId, stage, reason\nprocess_started\trunId, processId, context\nprocess_completed\trunId, processId, durationMs\nfeedback_received\ttaskId, agent, feedback\nCustomization\nAdd New Agents\n\nEdit colony/agents.yaml:\n\nagents:\n  myagent:\n    role: specialist\n    description: >\n      What this agent does...\n    model: anthropic/claude-sonnet-4\n    triggers:\n      - keyword1\n      - keyword2\n\n\nAfter adding, create their memory file:\n\ntouch colony/memory/myagent.md\n\nAdd New Processes\n\nEdit colony/processes.yaml:\n\nprocesses:\n  my-process:\n    description: \"What this process does\"\n    triggers: [keyword1, keyword2]\n    stages:\n      - id: stage1\n        agent: scout\n        task: \"Do something with: {context}\"\n        outputs: [output1.md]\n      - id: stage2\n        agent: pincer\n        task: \"Next step based on previous\"\n        inputs: [output1.md]\n        outputs: [output2.md]\n    checkpoints: [stage1]  # Optional: pause after these stages\n\nIntegration\n\nWorks with OpenClaw's agent sessions.\n\nDispatch/Assign (async): Tasks are spawned in the background and the CLI returns immediately. Use colony status to monitor progress and colony results <task-id> to view output.\n\nProcess stages (blocking): Multi-stage processes run sequentially, waiting for each stage to complete before proceeding. This ensures proper data flow between stages and checkpoint handling.\n\nEach agent receives:\n\nTheir role description\nLessons from their memory file\nActive facts from global context\nProject/preference context\nExamples\nValidate a Startup Idea\nnode scripts/colony.mjs process validate-idea \\\n  --context \"Subscription box for home coffee brewing experiments\"\n\n\nWatch as it flows: brainstorm → research → analyze → (checkpoint) → spec → estimate\n\nWrite and Publish a Blog Post\nnode scripts/colony.mjs process content-pipeline \\\n  --context \"Why RAG is eating traditional search\"\n\n\nStages: research → draft → (human review) → publish → promote\n\nQuick Research Task\nnode scripts/colony.mjs dispatch \"compare Pinecone vs Weaviate vs Milvus\"\n\n\nAuto-routes to scout, returns comparison.\n\nTrack Performance\n# After several tasks, check overall health\nnode scripts/colony.mjs audit\n\n# Deep dive into a struggling agent\nnode scripts/colony.mjs audit agent pincer\nnode scripts/colony.mjs audit failures\n\n# Add learnings from issues\nnode scripts/colony.mjs memory pincer add \"Handle file not found errors gracefully\" --mistake"
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/dabruhce/multi-agent-orchestration",
    "publisherUrl": "https://clawhub.ai/dabruhce/multi-agent-orchestration",
    "owner": "dabruhce",
    "version": "1.0.0",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/multi-agent-orchestration",
    "downloadUrl": "https://openagent3.xyz/downloads/multi-agent-orchestration",
    "agentUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/agent",
    "manifestUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/multi-agent-orchestration/agent.md"
  }
}