{
  "schemaVersion": "1.0",
  "item": {
    "slug": "lb-bmad-skill",
    "name": "BMad Method",
    "source": "tencent",
    "type": "skill",
    "category": "效率提升",
    "sourceUrl": "https://clawhub.ai/leonaaardob/lb-bmad-skill",
    "canonicalUrl": "https://clawhub.ai/leonaaardob/lb-bmad-skill",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/lb-bmad-skill",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=lb-bmad-skill",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "README.md",
      "SKILL.md",
      "docs/404.md",
      "docs/_STYLE_GUIDE.md",
      "docs/explanation/advanced-elicitation.md",
      "docs/explanation/adversarial-review.md"
    ],
    "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. Then review README.md for any prerequisites, environment setup, or post-install checks. 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. Then review README.md for any prerequisites, environment setup, or post-install checks. 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/lb-bmad-skill"
    },
    "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/lb-bmad-skill",
    "agentPageUrl": "https://openagent3.xyz/skills/lb-bmad-skill/agent",
    "manifestUrl": "https://openagent3.xyz/skills/lb-bmad-skill/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/lb-bmad-skill/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. Then review README.md for any prerequisites, environment setup, or post-install checks. 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. Then review README.md for any prerequisites, environment setup, or post-install checks. Summarize what changed and any follow-up checks I should run."
      }
    ]
  },
  "documentation": {
    "source": "clawhub",
    "primaryDoc": "SKILL.md",
    "sections": [
      {
        "title": "BMad Method Skill",
        "body": "Use BMad framework for AI-driven development with autonomous agent workflows.\n\nFor detailed reference, see:\n\ndocs/reference/commands.md - Complete command reference\ndocs/reference/agents.md - Available agents\ndocs/how-to/install-bmad.md - Detailed installation guide\ndocs/tutorials/getting-started.md - Quick start"
      },
      {
        "title": "DEPENDENCY",
        "body": "This skill requires coding-agent skill with Claude Code installed.\n\nClaude Code must be installed (~/.local/bin/claude)\nUse bash pty:true for all Claude Code invocations"
      },
      {
        "title": "Description",
        "body": "BMad (Breakthrough Method of Agile AI Driven Development) is a 4-phase framework:\n\nAnalysis — Explore the problem space\nPlanning — Define what to build\nSolutioning — Decide how to build it\nImplementation — Build it\n\nEach phase produces documents that become context for the next phase."
      },
      {
        "title": "Installation",
        "body": "To use BMad in a project:\n\n⚠️ Security Note: npx bmad-method install fetches code from npm. Only run this if you trust the BMad package. Review the package before installing.\n\ncd ~/project && npx bmad-method install\n\nSelect Claude Code when prompted."
      },
      {
        "title": "⚠️ Installation is Interactive",
        "body": "⚠️ npx bmad-method install asks questions!\n\nFor installation:\n\nDO NOT use background:true - you need to respond to prompts\nStay in the session and answer each question\nMonitor the log for these common prompts:\n\nPrompt in LogExpected AnswerNotes\"Where should BMad be installed?\". or ~path/to/projectCurrent directory\"Which AI tool are you using?\"Claude Code or numberSelect Claude\"Select modules to install\"a or enterSelect all/default\"Install BMad in current directory?\"y or enterConfirm\n\n# Installation must be interactive!\nbash pty:true workdir:~/project command:\"cd ~/project && npx bmad-method install\"\n# Stay present, answer each prompt:\n# - Monitor log for prompts\n# - Submit answer via: process action:submit sessionId:XXX data:\"y\""
      },
      {
        "title": "⚠️ Pre-Flight Check",
        "body": "Before running any /bmad- command, verify BMad is installed:\n\nls -la ~/project/_bmad/  # or _bmad-output/\n\nIf not found → run installation first:\n\nbash pty:true workdir:~/project command:\"cd ~/project && npx bmad-method install\""
      },
      {
        "title": "Model Selection",
        "body": "Strategic model selection for efficiency:\n\nModelBest Use CasesSonnetArchitecture, Solutioning, Quick-dev (complex tasks)HaikuBrainstorming, Story generation, Code review (repetitive/structured)OpusLarge refactoring, complex architecture decisions\n\n# Examples\nclaude --model sonnet \"Create the architecture\"\nclaude --model haiku \"Generate stories from the epic\""
      },
      {
        "title": "Available Commands (via /bmad-)",
        "body": "CommandPurposeOutput/bmad-helpInteractive guide-/bmad-brainstormingBrainstorm project ideas (use sparingly - see Notes)brainstorming-report.md/bmad-bmm-create-prdDefine requirementsPRD.md/bmad-bmm-create-ux-designDesign UXux-spec.md/bmad-bmm-create-architectureTechnical decisionsarchitecture.md + ADRs/bmad-bmm-create-epics-and-storiesBreak into storiesEpic files in _bmad-output//bmad-bmm-check-implementation-readinessGate checkPASS/CONCERNS/FAIL/bmad-bmm-sprint-planningInitialize sprintsprint-status.yaml/bmad-bmm-create-storyPrepare next storystory-[slug].md/bmad-bmm-dev-storyImplement storyWorking code + tests/bmad-bmm-code-reviewValidate qualityApproved/changes requested/bmad-bmm-quick-specQuick spec (skip phases 1-3)tech-spec.md/bmad-bmm-quick-devQuick implementationWorking code"
      },
      {
        "title": "Use Non-Interactive Mode When Possible",
        "body": "For commands that don't need real-time interaction:\n\n# Non-interactive (recommended for most BMad workflows)\nclaude -p --dangerously-skip-permissions \"Your prompt\""
      },
      {
        "title": "When to Use Background Mode",
        "body": "Use background:true only when:\n\nRunning multiple BMad workflows in sequence\nThe workflow is expected to take a long time\n\nAlways monitor with process action:log every 10-30 seconds to detect if Claude Code is waiting for input."
      },
      {
        "title": "Permission Configuration",
        "body": "To avoid Claude Code blocking on permission requests:\n\n⚠️ Security Note: Using --dangerously-skip-permissions or --permission-mode bypassPermissions suppresses permission checks. Use with caution - only for trusted code execution. For production workflows, prefer default permissions or validate the code first.\n\n# Skip all permission prompts (use with caution!)\nclaude --dangerously-skip-permissions \"prompt\"\n\n# Or use specific permission mode\nclaude --permission-mode bypassPermissions \"prompt\""
      },
      {
        "title": "Permission Loop Detection",
        "body": "If Claude Code waits for confirmation (Y/n, Commit, etc.):\n\nCheck the log: process action:log sessionId:XXX\nIdentify the type of prompt:\n\nShell command (Y/n): → submit \"y\"\nGit commit proposal: → submit \"n\" (see below)\nOther: → evaluate if you know the answer, otherwise ask user"
      },
      {
        "title": "Task Completion Detection (Background Mode)",
        "body": "How to know Claude Code is really done:\n\nSuccess message in log: Look for \"Task completed\", \"Done!\", \"All tasks finished\"\nPrompt available: The command prompt is back\nTimeout: If log is silent for 2+ minutes without completion message → check process:\nps aux | grep claude\nprocess action:log sessionId:XXX\n\n⚠️ Only consider task complete when you see explicit success message or prompt is back."
      },
      {
        "title": "Session Heartbeat (Long Running Tasks)",
        "body": "For workflows lasting 5+ minutes (/bmad-bmm-dev-story, large refactoring):\n\nEvery 60 seconds with no new log output:\n\n# Check if process is still alive\nps aux | grep claude\n\n# If stalled but alive → check if waiting for input\nprocess action:log sessionId:XXX\n\n# If process died → trigger recovery (see below)"
      },
      {
        "title": "Pattern 1: Full Analysis + Planning Request",
        "body": "User says: \"Analyze the current architecture and generate the product brief for project X\"\n\nAgent should:\n\nPre-flight check: Verify BMad installed (ls _bmad/)\nCheck project-context.md: If absent or outdated, generate it first (see below)\nLaunch Claude Code in the project directory:\nbash pty:true workdir:~/path/to/project background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'\"\n\n\nMonitor progress with process action:log (check every 10-30s)\nIf Claude Code needs information → ask the user directly\nWhen complete → run: ls _bmad-output/ to confirm files generated\nVerify output: grep -i \"error\" _bmad-output/architecture.md || head -20 _bmad-output/architecture.md\nRead architecture.md to verify coherence with user's request\nThen launch product brief: /bmad-bmm-create-product-brief"
      },
      {
        "title": "Pattern 2: Sprint Preparation with Story Generation",
        "body": "User says: \"Prepare sprint 1 and add tasks to OCM (OpenClaw Mission Center)\"\n\nAgent should:\n\nPre-flight check: Verify BMad installed\nLaunch Claude Code:\nbash pty:true workdir:~/path/to/project background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'\"\n\n\nWait for stories to be generated in _bmad-output/epics/\nRefresh context: run ls -R _bmad-output/ to confirm files exist\nRead stories efficiently (see \"Reading Stories Safely\" below)\nCreate OCM tasks from each story (use task-manager skill)\nReport completion with task list\n\nOCM Task Traceability\n\nWhen creating OCM tasks, ALWAYS include the BMad story reference:\n\nTask: Implement login form validation\nDescription: [full story content]\n---\nRef: _bmad-output/epics/auth/stories/story-login-validation.md\nEpic: auth\nCreated from: BMad Sprint 1\n\nWhy: This links the OCM task back to the source story for traceability."
      },
      {
        "title": "Pattern 3: Quick Feature Implementation",
        "body": "User says: \"Implement feature X using quick-dev\"\n\nAgent should:\n\nLaunch Claude Code with quick-dev:\nbash pty:true workdir:~/path/to/project command:\"claude --dangerously-skip-permissions '/bmad-bmm-quick-dev [feature description]'\""
      },
      {
        "title": "⚠️ Quick-Dev vs Standard: The Red Line",
        "body": "Quick-dev is faster but lacks safeguards. Use wisely.\n\n✅ OK with Quick-Dev❌ NEVER with Quick-DevUI tweaksAuthentication changesBug fixesEncryption/SecurityNew endpointsDatabase migrationsSimple featuresPayment logicBreaking schema changes\n\nRule: If the change touches security, auth, encryption, or database migrations → Use full BMad cycle (Analysis → Solutioning → Implementation)"
      },
      {
        "title": "project-context.md Management",
        "body": "BMad relies on project-context.md as the project's \"brain\". It's the persistent context that guides all decisions."
      },
      {
        "title": "Before /bmad-bmm-create-prd",
        "body": "Always check:\n\nls ~/project/project-context.md"
      },
      {
        "title": "If Missing or Outdated",
        "body": "Generate or update it:\n\n# Option 1: Generate from codebase\nbash pty:true workdir:~/project command:\"claude '/bmad-bmm-generate-project-context'\"\n\n# Option 2: Update manually with user's latest direction\n# Ask user: \"What's the current vision for this project?\"\n# Then create/update project-context.md with that info"
      },
      {
        "title": "When User Changes Direction",
        "body": "If user pivots mid-project (new features, different direction):\n\nUpdate project-context.md with new intentions\nRegenerate architecture.md if architecture is affected\nProceed with updated context"
      },
      {
        "title": "Reading Stories Safely (Avoid Context Overflow)",
        "body": "Don't dump all stories at once! Follow this process:\n\nList first:\nls _bmad-output/epics/*/stories/\n\n\n\nCheck each story header before reading full:\nhead -10 _bmad-output/epics/*/stories/story-*.md\n\n\n\nRead one at a time for task creation:\n\nRead story 1 → create OCM task\nRead story 2 → create OCM task\netc.\n\n\n\nFor batch operations, group by epic:\nfor f in _bmad-output/epics/*/stories/*.md; do head -20 \"$f\"; done | head -100"
      },
      {
        "title": "❌ Avoid This (Silent Failures)",
        "body": "claude \"/bmad-cmd1 && /bmad-cmd2\"  # If cmd1 fails, cmd2 still runs"
      },
      {
        "title": "✅ Prefer Sequential Execution",
        "body": "# Step 1: Run cmd1\nbash pty:true background:true command:\"claude '/bmad-cmd1'\"\n\n# Step 2: Verify output exists\nls _bmad-output/required-file.md\ngrep -q \"expected content\" _bmad-output/required-file.md || { echo \"FAILED\"; exit 1; }\n\n# Step 3: Run cmd2 only if cmd1 succeeded\nbash pty:true command:\"claude '/bmad-cmd2'\"\n\nRule: Verify each step before proceeding to the next."
      },
      {
        "title": "Recovery After Crash",
        "body": "Scenario: Claude Code crashes (API error 500, timeout, killed process)"
      },
      {
        "title": "Step 0: Kill Zombie Processes (BEFORE Restart!)",
        "body": "⚠️ Always check for stale processes first:\n\n# Check if Claude is still running\nps aux | grep claude\n\n# Kill any zombie processes for this project\npkill -f \"claude.*projects/roundvision\" || echo \"No zombie processes\"\n\n# Also kill any hanging node processes\npkill -f \"npx.*bmad\" || echo \"No zombie npx\""
      },
      {
        "title": "Step 1: Check what was generated",
        "body": "Check what was generated:\nls -lt _bmad-output/*.md | head -10\n\n\n\nFind the last valid file:\n# Read the most recently modified output\nls -t _bmad-output/*.md | head -1 | xargs head -30\n\n\n\nResume from where it stopped:\n\nIf architecture.md exists but stories/ missing → run story generation\nIf stories/ exist but no OCM tasks → create tasks from existing stories\nIf partial output → check coherence, regenerate only what's missing\n\n\n\nNever restart from zero if partial output exists"
      },
      {
        "title": "Handling Claude Code Questions",
        "body": "When Claude Code asks questions during execution:\n\nCheck the log first with process action:log sessionId:XXX to see what it asked\nIf you know the answer → provide it via process action:submit\nIf you need to ask the user → pause and get clarification first\nIf Claude Code is blocked → tell it to ask for what it needs, then come back to you\n\nExample:\n\n# Claude asks: \"What's your preferred authentication provider?\"\n# If you don't know → ask user: \"Claude needs to know auth provider - Auth0, Firebase, or Supabase?\"\n\n# Then provide the answer:\nprocess action:submit sessionId:XXX data:\"Auth0\""
      },
      {
        "title": "Use BMad for:",
        "body": "New features or epics\nArchitecture changes or refactoring\nSprint planning with story generation\nTechnical documentation (PRD, architecture)\nAnything security-sensitive"
      },
      {
        "title": "Use coding-agent directly (without BMad) for:",
        "body": "Quick fixes and small corrections\nSimple code reviews\nOne-file changes\nExperiments/prototyping\n\nRule of thumb: If it needs a story breakdown and sprint planning → BMad. If it's a simple edit → coding-agent directly."
      },
      {
        "title": "Reading BMad Outputs",
        "body": "After BMad workflows complete, documents are in:\n\nproject/\n├── _bmad/\n│   └── config.yaml\n├── _bmad-output/\n│   ├── brainstorming-report.md\n│   ├── product-brief.md\n│   ├── PRD.md\n│   ├── ux-spec.md\n│   ├── architecture.md\n│   ├── epics/\n│   │   └── epic-[name]/\n│   │       └── stories/\n│   │           └── story-[slug].md\n│   └── sprint-status.yaml\n└── project-context.md\n\n⚠️ Always verify files exist by running ls _bmad-output/ or ls -R _bmad-output/ after each workflow.\n\nVerify output validity before reading:\n\n# Quick check\nls -la _bmad-output/architecture.md\n\n# Validate content\nhead -20 _bmad-output/architecture.md\n\n# Check for errors\ngrep -i \"error\\|fail\" _bmad-output/architecture.md"
      },
      {
        "title": "Cache Refresh (Perception Reset)",
        "body": "⚠️ After Claude Code modifies source files, your cached view is stale!\n\nRule: After each successful Claude Code intervention on source code:\n\nDon't assume your previous read of a file is still valid\nRe-read the file if you need to work on it further\nClear mental cache - explicitly read the file again\n\n# Bad: Assuming old read is still valid\nread path:\"~/project/src/auth.js\"  # ❌ May be outdated\n\n# Good: Read fresh after Claude modified it\nexec command:\"cat ~/project/src/auth.js\"  # ✅ Fresh content"
      },
      {
        "title": "Validation Step",
        "body": "Before moving to Implementation phase:\n\nRead the generated architecture.md (or tech-spec.md for quick-dev)\nVerify it aligns with user's original request\nIf misaligned → regenerate or clarify with user"
      },
      {
        "title": "Error Handling",
        "body": "ErrorSolutionCommand not foundCheck PATH: echo $PATH and which claudenpx: command not foundInstall Node.js 20+_bmad/ not foundRun npx bmad-method install firstClaude stuck on permissionUse --dangerously-skip-permissionsAPI 500 errorTrigger recovery (see \"Recovery After Crash\")Session timeoutCheck if process still running, resume if possible"
      },
      {
        "title": "⚠️ Safety Rules",
        "body": "Never run rm -rf via Claude Code without explicit human validation\nNever use quick-dev for security-sensitive changes\nDefault Git answer: \"n\" (let OpenClaw handle commits)"
      },
      {
        "title": "Git Commit Handling",
        "body": "Claude Code often asks: \"Do you want to commit these changes? [y/N]\"\n\nReply \"n\" to keep Git control with OpenClaw\nReply \"y\" ONLY if user explicitly requested full Git autonomy\n\n# When Claude asks to commit, default to \"n\"\nprocess action:submit sessionId:XXX data:\"n\""
      },
      {
        "title": "Example 1: Architecture Analysis + Product Brief (Sequential)",
        "body": "User: \"On project PingRoot, analyze the current architecture and generate the product brief\"\n\nAgent does:\n\n# 1. Pre-flight check\nls ~/projects/pingroot/_bmad/ || echo \"Need to install BMad\"\n\n# 2. Check/update project-context.md\nls ~/projects/pingroot/project-context.md || echo \"Need to create project-context.md\"\n\n# 3. Launch architecture workflow\nbash pty:true workdir:~/projects/pingroot background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'\"\n\n# 4. Monitor, wait for completion\nprocess action:poll sessionId:XXX\n\n# 5. Verify output\nls _bmad-output/architecture.md\nhead -20 _bmad-output/architecture.md\ngrep -i \"error\" _bmad-output/architecture.md || echo \"OK\"\n\n# 6. If OK, verify coherence with user request\n# 7. If coherent, launch product brief\nbash pty:true workdir:~/projects/pingroot command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-product-brief'\"\n\n# 8. Deliver outputs"
      },
      {
        "title": "Example 2: Sprint Preparation + OCM Tasks (with safety checks)",
        "body": "User: \"Prepare sprint 1 for RoundVision and add tasks to OCM\"\n\nAgent does:\n\n# 1. Pre-flight check\nls ~/projects/roundvision/_bmad/ || npx bmad-method install\n\n# 2. Check project-context.md\nls ~/projects/roundvision/project-context.md || echo \"Update this first!\"\n\n# 3. Launch sprint planning + story creation\nbash pty:true workdir:~/projects/roundvision background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'\"\n\n# 4. Monitor and wait for completion\nprocess action:poll sessionId:XXX  # repeat until done\n\n# 5. Refresh context - verify files exist\nls -R ~/projects/roundvision/_bmad-output/epics/\n\n# 6. List stories first (don't dump all at once!)\nls ~/projects/roundvision/_bmad-output/epics/*/stories/\n\n# 7. Read and process stories one by one\nfor story in ~/projects/roundvision/_bmad-output/epics/*/stories/*.md; do\n  echo \"=== $(basename $story) ===\"\n  head -20 \"$story\"\n  # Create OCM task from this story\ndone\n\n# 8. Report: \"Created X tasks in OCM for Sprint 1\"\n#    IMPORTANT: Each OCM task must include story path as reference!"
      },
      {
        "title": "Example 3: Quick Fix (No BMad Needed)",
        "body": "User: \"Fix the typo in the login page\"\n\nAgent does:\n\n# Direct coding-agent, no BMad workflow needed\nbash pty:true workdir:~/projects/login command:\"claude 'Fix the typo on line 42: \\\"Passowrd\\\" → \\\"Password\\\"'\""
      },
      {
        "title": "Example 4: Recovery After Crash",
        "body": "Scenario: Claude Code crashes during story generation\n\nAgent does:\n\n# 0. Cleanup zombies FIRST!\nps aux | grep claude\npkill -f \"claude.*projects/roundvision\" || echo \"Clean\"\n\n# 1. Check what was generated\nls -lt ~/project/_bmad-output/ | head -10\n\n# 2. Find last valid file\nls -t ~/project/_bmad-output/epics/*/stories/ | head -1\n\n# 3. Check if partial stories exist\nls ~/project/_bmad-output/epics/*/stories/*.md | wc -l\n\n# 4. If partial → resume from last point\n# If 3 stories out of 5 → generate remaining 2\n# If 0 stories → restart story generation\n\n# 5. Continue without restarting from zero"
      },
      {
        "title": "⚠️ CRITICAL: Sub-Agent (Minion) Access",
        "body": "The minion does NOT automatically have access to project files! sub-agent to implement\n\nWhen spawning a a task, you MUST provide:"
      },
      {
        "title": "1. Project Directory Access",
        "body": "# Minion needs workdir to access project files\nsessions_spawn workdir:\"~/projects/roundvision\" ..."
      },
      {
        "title": "2. Story + Context + Architecture",
        "body": "⚠️ NEVER give only the story to a minion!\n\nThe story says \"Add a login button\" but doesn't say:\n\nIs this React, Vue, or vanilla JS?\nDoes it use Tailwind or Bootstrap?\nWhat's the existing auth pattern?\n\nYou MUST provide:\n\nStory (what to build)\nproject-context.md (project rules, tech stack)\narchitecture.md (technical decisions)\n\n# Step 1: Read all three\ncat ~/projects/roundvision/project-context.md\ncat ~/projects/roundvision/_bmad-output/architecture.md  \ncat ~/projects/roundvision/_bmad-output/epics/auth.md\n\n# Step/stories/story-login 2: Combine into a comprehensive prompt\nsessions_spawn task:\"You are implementing this story: [STORY]. \nProject context: [CONTEXT].\nArchitecture: [ARCHITECTURE].\nFollow the patterns defined in architecture.md.\""
      },
      {
        "title": "3. OCM Task Should Include Story Path",
        "body": "{\n  \"title\": \"Implement login form validation\",\n  \"description\": \"Full story content...\",\n  \"source\": \"_bmad-output/epics/auth/stories/story-login.md\"\n}\n\n⚠️ Without workdir + story content + context + architecture, the minion is blind and cannot implement anything!"
      },
      {
        "title": "Notes",
        "body": "BMad brainstorming: Use sparingly. OpenClaw itself is a brainstorming agent. Use BMad for technical structuring, keep high-level strategy with OpenClaw.\nBMad generates files in _bmad-output/\nproject-context.md is the project's brain - keep it updated\n/bmad-help provides interactive guidance\nAlways use pty:true with Claude Code\nFor model restrictions, use --model <sonnet|haiku|opus>\nToken efficiency: Use direct coding-agent for small tasks, reserve BMad for complex workflows\nSequential over chained: Verify each step before proceeding"
      },
      {
        "title": "Related Skills",
        "body": "coding-agent — Required for launching Claude Code\ntask-manager — For creating OCM tasks from BMad stories"
      }
    ],
    "body": "BMad Method Skill\n\nUse BMad framework for AI-driven development with autonomous agent workflows.\n\nFor detailed reference, see:\n\ndocs/reference/commands.md - Complete command reference\ndocs/reference/agents.md - Available agents\ndocs/how-to/install-bmad.md - Detailed installation guide\ndocs/tutorials/getting-started.md - Quick start\nDEPENDENCY\n\nThis skill requires coding-agent skill with Claude Code installed.\n\nClaude Code must be installed (~/.local/bin/claude)\nUse bash pty:true for all Claude Code invocations\nDescription\n\nBMad (Breakthrough Method of Agile AI Driven Development) is a 4-phase framework:\n\nAnalysis — Explore the problem space\nPlanning — Define what to build\nSolutioning — Decide how to build it\nImplementation — Build it\n\nEach phase produces documents that become context for the next phase.\n\nInstallation\n\nTo use BMad in a project:\n\n⚠️ Security Note: npx bmad-method install fetches code from npm. Only run this if you trust the BMad package. Review the package before installing.\n\ncd ~/project && npx bmad-method install\n\n\nSelect Claude Code when prompted.\n\n⚠️ Installation is Interactive\n\n⚠️ npx bmad-method install asks questions!\n\nFor installation:\n\nDO NOT use background:true - you need to respond to prompts\nStay in the session and answer each question\nMonitor the log for these common prompts:\nPrompt in Log\tExpected Answer\tNotes\n\"Where should BMad be installed?\"\t. or ~path/to/project\tCurrent directory\n\"Which AI tool are you using?\"\tClaude Code or number\tSelect Claude\n\"Select modules to install\"\ta or enter\tSelect all/default\n\"Install BMad in current directory?\"\ty or enter\tConfirm\n# Installation must be interactive!\nbash pty:true workdir:~/project command:\"cd ~/project && npx bmad-method install\"\n# Stay present, answer each prompt:\n# - Monitor log for prompts\n# - Submit answer via: process action:submit sessionId:XXX data:\"y\"\n\n⚠️ Pre-Flight Check\n\nBefore running any /bmad- command, verify BMad is installed:\n\nls -la ~/project/_bmad/  # or _bmad-output/\n\n\nIf not found → run installation first:\n\nbash pty:true workdir:~/project command:\"cd ~/project && npx bmad-method install\"\n\nModel Selection\n\nStrategic model selection for efficiency:\n\nModel\tBest Use Cases\nSonnet\tArchitecture, Solutioning, Quick-dev (complex tasks)\nHaiku\tBrainstorming, Story generation, Code review (repetitive/structured)\nOpus\tLarge refactoring, complex architecture decisions\n# Examples\nclaude --model sonnet \"Create the architecture\"\nclaude --model haiku \"Generate stories from the epic\"\n\nAvailable Commands (via /bmad-)\nCommand\tPurpose\tOutput\n/bmad-help\tInteractive guide\t-\n/bmad-brainstorming\tBrainstorm project ideas (use sparingly - see Notes)\tbrainstorming-report.md\n/bmad-bmm-create-prd\tDefine requirements\tPRD.md\n/bmad-bmm-create-ux-design\tDesign UX\tux-spec.md\n/bmad-bmm-create-architecture\tTechnical decisions\tarchitecture.md + ADRs\n/bmad-bmm-create-epics-and-stories\tBreak into stories\tEpic files in _bmad-output/\n/bmad-bmm-check-implementation-readiness\tGate check\tPASS/CONCERNS/FAIL\n/bmad-bmm-sprint-planning\tInitialize sprint\tsprint-status.yaml\n/bmad-bmm-create-story\tPrepare next story\tstory-[slug].md\n/bmad-bmm-dev-story\tImplement story\tWorking code + tests\n/bmad-bmm-code-review\tValidate quality\tApproved/changes requested\n/bmad-bmm-quick-spec\tQuick spec (skip phases 1-3)\ttech-spec.md\n/bmad-bmm-quick-dev\tQuick implementation\tWorking code\n⚠️ Important: Claude Code Execution\nUse Non-Interactive Mode When Possible\n\nFor commands that don't need real-time interaction:\n\n# Non-interactive (recommended for most BMad workflows)\nclaude -p --dangerously-skip-permissions \"Your prompt\"\n\nWhen to Use Background Mode\n\nUse background:true only when:\n\nRunning multiple BMad workflows in sequence\nThe workflow is expected to take a long time\n\nAlways monitor with process action:log every 10-30 seconds to detect if Claude Code is waiting for input.\n\nPermission Configuration\n\nTo avoid Claude Code blocking on permission requests:\n\n⚠️ Security Note: Using --dangerously-skip-permissions or --permission-mode bypassPermissions suppresses permission checks. Use with caution - only for trusted code execution. For production workflows, prefer default permissions or validate the code first.\n\n# Skip all permission prompts (use with caution!)\nclaude --dangerously-skip-permissions \"prompt\"\n\n# Or use specific permission mode\nclaude --permission-mode bypassPermissions \"prompt\"\n\nPermission Loop Detection\n\nIf Claude Code waits for confirmation (Y/n, Commit, etc.):\n\nCheck the log: process action:log sessionId:XXX\nIdentify the type of prompt:\nShell command (Y/n): → submit \"y\"\nGit commit proposal: → submit \"n\" (see below)\nOther: → evaluate if you know the answer, otherwise ask user\nTask Completion Detection (Background Mode)\n\nHow to know Claude Code is really done:\n\nSuccess message in log: Look for \"Task completed\", \"Done!\", \"All tasks finished\"\nPrompt available: The command prompt is back\nTimeout: If log is silent for 2+ minutes without completion message → check process:\nps aux | grep claude\nprocess action:log sessionId:XXX\n\n\n⚠️ Only consider task complete when you see explicit success message or prompt is back.\n\nSession Heartbeat (Long Running Tasks)\n\nFor workflows lasting 5+ minutes (/bmad-bmm-dev-story, large refactoring):\n\nEvery 60 seconds with no new log output:\n\n# Check if process is still alive\nps aux | grep claude\n\n# If stalled but alive → check if waiting for input\nprocess action:log sessionId:XXX\n\n# If process died → trigger recovery (see below)\n\nAutonomous Workflow Patterns\nPattern 1: Full Analysis + Planning Request\n\nUser says: \"Analyze the current architecture and generate the product brief for project X\"\n\nAgent should:\n\nPre-flight check: Verify BMad installed (ls _bmad/)\nCheck project-context.md: If absent or outdated, generate it first (see below)\nLaunch Claude Code in the project directory:\nbash pty:true workdir:~/path/to/project background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'\"\n\nMonitor progress with process action:log (check every 10-30s)\nIf Claude Code needs information → ask the user directly\nWhen complete → run: ls _bmad-output/ to confirm files generated\nVerify output: grep -i \"error\" _bmad-output/architecture.md || head -20 _bmad-output/architecture.md\nRead architecture.md to verify coherence with user's request\nThen launch product brief: /bmad-bmm-create-product-brief\nPattern 2: Sprint Preparation with Story Generation\n\nUser says: \"Prepare sprint 1 and add tasks to OCM (OpenClaw Mission Center)\"\n\nAgent should:\n\nPre-flight check: Verify BMad installed\nLaunch Claude Code:\nbash pty:true workdir:~/path/to/project background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'\"\n\nWait for stories to be generated in _bmad-output/epics/\nRefresh context: run ls -R _bmad-output/ to confirm files exist\nRead stories efficiently (see \"Reading Stories Safely\" below)\nCreate OCM tasks from each story (use task-manager skill)\nReport completion with task list\nOCM Task Traceability\n\nWhen creating OCM tasks, ALWAYS include the BMad story reference:\n\nTask: Implement login form validation\nDescription: [full story content]\n---\nRef: _bmad-output/epics/auth/stories/story-login-validation.md\nEpic: auth\nCreated from: BMad Sprint 1\n\n\nWhy: This links the OCM task back to the source story for traceability.\n\nPattern 3: Quick Feature Implementation\n\nUser says: \"Implement feature X using quick-dev\"\n\nAgent should:\n\nLaunch Claude Code with quick-dev:\nbash pty:true workdir:~/path/to/project command:\"claude --dangerously-skip-permissions '/bmad-bmm-quick-dev [feature description]'\"\n\n⚠️ Quick-Dev vs Standard: The Red Line\n\nQuick-dev is faster but lacks safeguards. Use wisely.\n\n✅ OK with Quick-Dev\t❌ NEVER with Quick-Dev\nUI tweaks\tAuthentication changes\nBug fixes\tEncryption/Security\nNew endpoints\tDatabase migrations\nSimple features\tPayment logic\n\tBreaking schema changes\n\nRule: If the change touches security, auth, encryption, or database migrations → Use full BMad cycle (Analysis → Solutioning → Implementation)\n\nproject-context.md Management\n\nBMad relies on project-context.md as the project's \"brain\". It's the persistent context that guides all decisions.\n\nBefore /bmad-bmm-create-prd\n\nAlways check:\n\nls ~/project/project-context.md\n\nIf Missing or Outdated\n\nGenerate or update it:\n\n# Option 1: Generate from codebase\nbash pty:true workdir:~/project command:\"claude '/bmad-bmm-generate-project-context'\"\n\n# Option 2: Update manually with user's latest direction\n# Ask user: \"What's the current vision for this project?\"\n# Then create/update project-context.md with that info\n\nWhen User Changes Direction\n\nIf user pivots mid-project (new features, different direction):\n\nUpdate project-context.md with new intentions\nRegenerate architecture.md if architecture is affected\nProceed with updated context\nReading Stories Safely (Avoid Context Overflow)\n\nDon't dump all stories at once! Follow this process:\n\nList first:\n\nls _bmad-output/epics/*/stories/\n\n\nCheck each story header before reading full:\n\nhead -10 _bmad-output/epics/*/stories/story-*.md\n\n\nRead one at a time for task creation:\n\nRead story 1 → create OCM task\nRead story 2 → create OCM task\netc.\n\nFor batch operations, group by epic:\n\nfor f in _bmad-output/epics/*/stories/*.md; do head -20 \"$f\"; done | head -100\n\nCommand Chain Safety\n❌ Avoid This (Silent Failures)\nclaude \"/bmad-cmd1 && /bmad-cmd2\"  # If cmd1 fails, cmd2 still runs\n\n✅ Prefer Sequential Execution\n# Step 1: Run cmd1\nbash pty:true background:true command:\"claude '/bmad-cmd1'\"\n\n# Step 2: Verify output exists\nls _bmad-output/required-file.md\ngrep -q \"expected content\" _bmad-output/required-file.md || { echo \"FAILED\"; exit 1; }\n\n# Step 3: Run cmd2 only if cmd1 succeeded\nbash pty:true command:\"claude '/bmad-cmd2'\"\n\n\nRule: Verify each step before proceeding to the next.\n\nRecovery After Crash\n\nScenario: Claude Code crashes (API error 500, timeout, killed process)\n\nStep 0: Kill Zombie Processes (BEFORE Restart!)\n\n⚠️ Always check for stale processes first:\n\n# Check if Claude is still running\nps aux | grep claude\n\n# Kill any zombie processes for this project\npkill -f \"claude.*projects/roundvision\" || echo \"No zombie processes\"\n\n# Also kill any hanging node processes\npkill -f \"npx.*bmad\" || echo \"No zombie npx\"\n\nStep 1: Check what was generated\n\nCheck what was generated:\n\nls -lt _bmad-output/*.md | head -10\n\n\nFind the last valid file:\n\n# Read the most recently modified output\nls -t _bmad-output/*.md | head -1 | xargs head -30\n\n\nResume from where it stopped:\n\nIf architecture.md exists but stories/ missing → run story generation\nIf stories/ exist but no OCM tasks → create tasks from existing stories\nIf partial output → check coherence, regenerate only what's missing\n\nNever restart from zero if partial output exists\n\nHandling Claude Code Questions\n\nWhen Claude Code asks questions during execution:\n\nCheck the log first with process action:log sessionId:XXX to see what it asked\nIf you know the answer → provide it via process action:submit\nIf you need to ask the user → pause and get clarification first\nIf Claude Code is blocked → tell it to ask for what it needs, then come back to you\n\nExample:\n\n# Claude asks: \"What's your preferred authentication provider?\"\n# If you don't know → ask user: \"Claude needs to know auth provider - Auth0, Firebase, or Supabase?\"\n\n# Then provide the answer:\nprocess action:submit sessionId:XXX data:\"Auth0\"\n\nWhen to Use BMad vs Direct Coding-Agent\nUse BMad for:\nNew features or epics\nArchitecture changes or refactoring\nSprint planning with story generation\nTechnical documentation (PRD, architecture)\nAnything security-sensitive\nUse coding-agent directly (without BMad) for:\nQuick fixes and small corrections\nSimple code reviews\nOne-file changes\nExperiments/prototyping\n\nRule of thumb: If it needs a story breakdown and sprint planning → BMad. If it's a simple edit → coding-agent directly.\n\nReading BMad Outputs\n\nAfter BMad workflows complete, documents are in:\n\nproject/\n├── _bmad/\n│   └── config.yaml\n├── _bmad-output/\n│   ├── brainstorming-report.md\n│   ├── product-brief.md\n│   ├── PRD.md\n│   ├── ux-spec.md\n│   ├── architecture.md\n│   ├── epics/\n│   │   └── epic-[name]/\n│   │       └── stories/\n│   │           └── story-[slug].md\n│   └── sprint-status.yaml\n└── project-context.md\n\n\n⚠️ Always verify files exist by running ls _bmad-output/ or ls -R _bmad-output/ after each workflow.\n\nVerify output validity before reading:\n\n# Quick check\nls -la _bmad-output/architecture.md\n\n# Validate content\nhead -20 _bmad-output/architecture.md\n\n# Check for errors\ngrep -i \"error\\|fail\" _bmad-output/architecture.md\n\nCache Refresh (Perception Reset)\n\n⚠️ After Claude Code modifies source files, your cached view is stale!\n\nRule: After each successful Claude Code intervention on source code:\n\nDon't assume your previous read of a file is still valid\nRe-read the file if you need to work on it further\nClear mental cache - explicitly read the file again\n# Bad: Assuming old read is still valid\nread path:\"~/project/src/auth.js\"  # ❌ May be outdated\n\n# Good: Read fresh after Claude modified it\nexec command:\"cat ~/project/src/auth.js\"  # ✅ Fresh content\n\nValidation Step\n\nBefore moving to Implementation phase:\n\nRead the generated architecture.md (or tech-spec.md for quick-dev)\nVerify it aligns with user's original request\nIf misaligned → regenerate or clarify with user\nError Handling\nError\tSolution\nCommand not found\tCheck PATH: echo $PATH and which claude\nnpx: command not found\tInstall Node.js 20+\n_bmad/ not found\tRun npx bmad-method install first\nClaude stuck on permission\tUse --dangerously-skip-permissions\nAPI 500 error\tTrigger recovery (see \"Recovery After Crash\")\nSession timeout\tCheck if process still running, resume if possible\n⚠️ Safety Rules\nNever run rm -rf via Claude Code without explicit human validation\nNever use quick-dev for security-sensitive changes\nDefault Git answer: \"n\" (let OpenClaw handle commits)\nGit Commit Handling\n\nClaude Code often asks: \"Do you want to commit these changes? [y/N]\"\n\nReply \"n\" to keep Git control with OpenClaw\nReply \"y\" ONLY if user explicitly requested full Git autonomy\n# When Claude asks to commit, default to \"n\"\nprocess action:submit sessionId:XXX data:\"n\"\n\nExamples\nExample 1: Architecture Analysis + Product Brief (Sequential)\n\nUser: \"On project PingRoot, analyze the current architecture and generate the product brief\"\n\nAgent does:\n\n# 1. Pre-flight check\nls ~/projects/pingroot/_bmad/ || echo \"Need to install BMad\"\n\n# 2. Check/update project-context.md\nls ~/projects/pingroot/project-context.md || echo \"Need to create project-context.md\"\n\n# 3. Launch architecture workflow\nbash pty:true workdir:~/projects/pingroot background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'\"\n\n# 4. Monitor, wait for completion\nprocess action:poll sessionId:XXX\n\n# 5. Verify output\nls _bmad-output/architecture.md\nhead -20 _bmad-output/architecture.md\ngrep -i \"error\" _bmad-output/architecture.md || echo \"OK\"\n\n# 6. If OK, verify coherence with user request\n# 7. If coherent, launch product brief\nbash pty:true workdir:~/projects/pingroot command:\"claude --dangerously-skip-permissions '/bmad-bmm-create-product-brief'\"\n\n# 8. Deliver outputs\n\nExample 2: Sprint Preparation + OCM Tasks (with safety checks)\n\nUser: \"Prepare sprint 1 for RoundVision and add tasks to OCM\"\n\nAgent does:\n\n# 1. Pre-flight check\nls ~/projects/roundvision/_bmad/ || npx bmad-method install\n\n# 2. Check project-context.md\nls ~/projects/roundvision/project-context.md || echo \"Update this first!\"\n\n# 3. Launch sprint planning + story creation\nbash pty:true workdir:~/projects/roundvision background:true command:\"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'\"\n\n# 4. Monitor and wait for completion\nprocess action:poll sessionId:XXX  # repeat until done\n\n# 5. Refresh context - verify files exist\nls -R ~/projects/roundvision/_bmad-output/epics/\n\n# 6. List stories first (don't dump all at once!)\nls ~/projects/roundvision/_bmad-output/epics/*/stories/\n\n# 7. Read and process stories one by one\nfor story in ~/projects/roundvision/_bmad-output/epics/*/stories/*.md; do\n  echo \"=== $(basename $story) ===\"\n  head -20 \"$story\"\n  # Create OCM task from this story\ndone\n\n# 8. Report: \"Created X tasks in OCM for Sprint 1\"\n#    IMPORTANT: Each OCM task must include story path as reference!\n\nExample 3: Quick Fix (No BMad Needed)\n\nUser: \"Fix the typo in the login page\"\n\nAgent does:\n\n# Direct coding-agent, no BMad workflow needed\nbash pty:true workdir:~/projects/login command:\"claude 'Fix the typo on line 42: \\\"Passowrd\\\" → \\\"Password\\\"'\"\n\nExample 4: Recovery After Crash\n\nScenario: Claude Code crashes during story generation\n\nAgent does:\n\n# 0. Cleanup zombies FIRST!\nps aux | grep claude\npkill -f \"claude.*projects/roundvision\" || echo \"Clean\"\n\n# 1. Check what was generated\nls -lt ~/project/_bmad-output/ | head -10\n\n# 2. Find last valid file\nls -t ~/project/_bmad-output/epics/*/stories/ | head -1\n\n# 3. Check if partial stories exist\nls ~/project/_bmad-output/epics/*/stories/*.md | wc -l\n\n# 4. If partial → resume from last point\n# If 3 stories out of 5 → generate remaining 2\n# If 0 stories → restart story generation\n\n# 5. Continue without restarting from zero\n\n⚠️ CRITICAL: Sub-Agent (Minion) Access\n\nThe minion does NOT automatically have access to project files! sub-agent to implement\n\nWhen spawning a a task, you MUST provide:\n\n1. Project Directory Access\n# Minion needs workdir to access project files\nsessions_spawn workdir:\"~/projects/roundvision\" ...\n\n2. Story + Context + Architecture\n\n⚠️ NEVER give only the story to a minion!\n\nThe story says \"Add a login button\" but doesn't say:\n\nIs this React, Vue, or vanilla JS?\nDoes it use Tailwind or Bootstrap?\nWhat's the existing auth pattern?\n\nYou MUST provide:\n\nStory (what to build)\nproject-context.md (project rules, tech stack)\narchitecture.md (technical decisions)\n# Step 1: Read all three\ncat ~/projects/roundvision/project-context.md\ncat ~/projects/roundvision/_bmad-output/architecture.md  \ncat ~/projects/roundvision/_bmad-output/epics/auth.md\n\n# Step/stories/story-login 2: Combine into a comprehensive prompt\nsessions_spawn task:\"You are implementing this story: [STORY]. \nProject context: [CONTEXT].\nArchitecture: [ARCHITECTURE].\nFollow the patterns defined in architecture.md.\"\n\n3. OCM Task Should Include Story Path\n{\n  \"title\": \"Implement login form validation\",\n  \"description\": \"Full story content...\",\n  \"source\": \"_bmad-output/epics/auth/stories/story-login.md\"\n}\n\n\n⚠️ Without workdir + story content + context + architecture, the minion is blind and cannot implement anything!\n\nNotes\nBMad brainstorming: Use sparingly. OpenClaw itself is a brainstorming agent. Use BMad for technical structuring, keep high-level strategy with OpenClaw.\nBMad generates files in _bmad-output/\nproject-context.md is the project's brain - keep it updated\n/bmad-help provides interactive guidance\nAlways use pty:true with Claude Code\nFor model restrictions, use --model <sonnet|haiku|opus>\nToken efficiency: Use direct coding-agent for small tasks, reserve BMad for complex workflows\nSequential over chained: Verify each step before proceeding\nRelated Skills\ncoding-agent — Required for launching Claude Code\ntask-manager — For creating OCM tasks from BMad stories"
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/leonaaardob/lb-bmad-skill",
    "publisherUrl": "https://clawhub.ai/leonaaardob/lb-bmad-skill",
    "owner": "leonaaardob",
    "version": "1.0.2",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/lb-bmad-skill",
    "downloadUrl": "https://openagent3.xyz/downloads/lb-bmad-skill",
    "agentUrl": "https://openagent3.xyz/skills/lb-bmad-skill/agent",
    "manifestUrl": "https://openagent3.xyz/skills/lb-bmad-skill/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/lb-bmad-skill/agent.md"
  }
}