菜单
Use BMad (Breakthrough Method of Agile AI Driven Development) framework for AI-driven development. Use for: architecture analysis, sprint planning, story generation, PRD creation, and full development workflows. Requires coding-agent skill with Claude Code.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/johnalbertini14-glitch/openclaw-skills --skill bmad-method命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use this skill to create a Polymarket wallet for your agent and trade on prediction markets. Browse markets, place bets, manage positions — all without exposing private keys.
ClawSec suite manager with embedded advisory-feed monitoring, cryptographic signature verification, approval-gated malicious-skill response, and guided setup for additional security skills.
Automated daily security audits for OpenClaw agents with email reporting. Runs deep audits and sends formatted reports.
基于 SOC 职业分类
正在显示 SKILL.md
| name | bmad-method |
| description | Use BMad (Breakthrough Method of Agile AI Driven Development) framework for AI-driven development. Use for: architecture analysis, sprint planning, story generation, PRD creation, and full development workflows. Requires coding-agent skill with Claude Code. |
Use BMad framework for AI-driven development with autonomous agent workflows.
For detailed reference, see:
This skill requires coding-agent skill with Claude Code installed.
~/.local/bin/claude)bash pty:true for all Claude Code invocationsBMad (Breakthrough Method of Agile AI Driven Development) is a 4-phase framework:
Each phase produces documents that become context for the next phase.
To use BMad in a project:
⚠️ Security Note:
npx bmad-method installfetches code from npm. Only run this if you trust the BMad package. Review the package before installing.
cd ~/project && npx bmad-method install
Select Claude Code when prompted.
⚠️ npx bmad-method install asks questions!
For installation:
| Prompt in Log | Expected Answer | Notes |
|---|---|---|
| "Where should BMad be installed?" | . or ~path/to/project | Current directory |
| "Which AI tool are you using?" | Claude Code or number | Select Claude |
| "Select modules to install" | a or enter | Select all/default |
| "Install BMad in current directory?" | y or enter | Confirm |
# Installation must be interactive!
bash pty:true workdir:~/project command:"cd ~/project && npx bmad-method install"
# Stay present, answer each prompt:
# - Monitor log for prompts
# - Submit answer via: process action:submit sessionId:XXX data:"y"
Before running any /bmad- command, verify BMad is installed:
ls -la ~/project/_bmad/ # or _bmad-output/
If not found → run installation first:
bash pty:true workdir:~/project command:"cd ~/project && npx bmad-method install"
Strategic model selection for efficiency:
| Model | Best Use Cases |
|---|---|
| Sonnet | Architecture, Solutioning, Quick-dev (complex tasks) |
| Haiku | Brainstorming, Story generation, Code review (repetitive/structured) |
| Opus | Large refactoring, complex architecture decisions |
# Examples
claude --model sonnet "Create the architecture"
claude --model haiku "Generate stories from the epic"
| Command | Purpose | Output |
|---|---|---|
/bmad-help | Interactive guide | - |
/bmad-brainstorming | Brainstorm project ideas (use sparingly - see Notes) | brainstorming-report.md |
/bmad-bmm-create-prd | Define requirements | PRD.md |
/bmad-bmm-create-ux-design | Design UX | ux-spec.md |
/bmad-bmm-create-architecture | Technical decisions | architecture.md + ADRs |
/bmad-bmm-create-epics-and-stories | Break into stories | Epic files in _bmad-output/ |
/bmad-bmm-check-implementation-readiness | Gate check | PASS/CONCERNS/FAIL |
/bmad-bmm-sprint-planning | Initialize sprint | sprint-status.yaml |
/bmad-bmm-create-story | Prepare next story | story-[slug].md |
/bmad-bmm-dev-story | Implement story | Working code + tests |
/bmad-bmm-code-review | Validate quality | Approved/changes requested |
/bmad-bmm-quick-spec | Quick spec (skip phases 1-3) | tech-spec.md |
/bmad-bmm-quick-dev | Quick implementation | Working code |
For commands that don't need real-time interaction:
# Non-interactive (recommended for most BMad workflows)
claude -p --dangerously-skip-permissions "Your prompt"
Use background:true only when:
Always monitor with process action:log every 10-30 seconds to detect if Claude Code is waiting for input.
To avoid Claude Code blocking on permission requests:
⚠️ Security Note: Using
--dangerously-skip-permissionsor--permission-mode bypassPermissionssuppresses permission checks. Use with caution - only for trusted code execution. For production workflows, prefer default permissions or validate the code first.
# Skip all permission prompts (use with caution!)
claude --dangerously-skip-permissions "prompt"
# Or use specific permission mode
claude --permission-mode bypassPermissions "prompt"
If Claude Code waits for confirmation (Y/n, Commit, etc.):
process action:log sessionId:XXXHow to know Claude Code is really done:
ps aux | grep claude
process action:log sessionId:XXX
⚠️ Only consider task complete when you see explicit success message or prompt is back.
For workflows lasting 5+ minutes (/bmad-bmm-dev-story, large refactoring):
Every 60 seconds with no new log output:
# Check if process is still alive
ps aux | grep claude
# If stalled but alive → check if waiting for input
process action:log sessionId:XXX
# If process died → trigger recovery (see below)
User says: "Analyze the current architecture and generate the product brief for project X"
Agent should:
ls _bmad/)bash pty:true workdir:~/path/to/project background:true command:"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'"
process action:log (check every 10-30s)ls _bmad-output/ to confirm files generatedgrep -i "error" _bmad-output/architecture.md || head -20 _bmad-output/architecture.mdarchitecture.md to verify coherence with user's request/bmad-bmm-create-product-briefUser says: "Prepare sprint 1 and add tasks to OCM (OpenClaw Mission Center)"
Agent should:
bash pty:true workdir:~/path/to/project background:true command:"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'"
_bmad-output/epics/ls -R _bmad-output/ to confirm files existWhen creating OCM tasks, ALWAYS include the BMad story reference:
Task: Implement login form validation
Description: [full story content]
---
Ref: _bmad-output/epics/auth/stories/story-login-validation.md
Epic: auth
Created from: BMad Sprint 1
Why: This links the OCM task back to the source story for traceability.
User says: "Implement feature X using quick-dev"
Agent should:
bash pty:true workdir:~/path/to/project command:"claude --dangerously-skip-permissions '/bmad-bmm-quick-dev [feature description]'"
Quick-dev is faster but lacks safeguards. Use wisely.
| ✅ OK with Quick-Dev | ❌ NEVER with Quick-Dev |
|---|---|
| UI tweaks | Authentication changes |
| Bug fixes | Encryption/Security |
| New endpoints | Database migrations |
| Simple features | Payment logic |
| Breaking schema changes |
Rule: If the change touches security, auth, encryption, or database migrations → Use full BMad cycle (Analysis → Solutioning → Implementation)
BMad relies on project-context.md as the project's "brain". It's the persistent context that guides all decisions.
Always check:
ls ~/project/project-context.md
Generate or update it:
# Option 1: Generate from codebase
bash pty:true workdir:~/project command:"claude '/bmad-bmm-generate-project-context'"
# Option 2: Update manually with user's latest direction
# Ask user: "What's the current vision for this project?"
# Then create/update project-context.md with that info
If user pivots mid-project (new features, different direction):
project-context.md with new intentionsarchitecture.md if architecture is affectedDon't dump all stories at once! Follow this process:
List first:
ls _bmad-output/epics/*/stories/
Check each story header before reading full:
head -10 _bmad-output/epics/*/stories/story-*.md
Read one at a time for task creation:
For batch operations, group by epic:
for f in _bmad-output/epics/*/stories/*.md; do head -20 "$f"; done | head -100
claude "/bmad-cmd1 && /bmad-cmd2" # If cmd1 fails, cmd2 still runs
# Step 1: Run cmd1
bash pty:true background:true command:"claude '/bmad-cmd1'"
# Step 2: Verify output exists
ls _bmad-output/required-file.md
grep -q "expected content" _bmad-output/required-file.md || { echo "FAILED"; exit 1; }
# Step 3: Run cmd2 only if cmd1 succeeded
bash pty:true command:"claude '/bmad-cmd2'"
Rule: Verify each step before proceeding to the next.
Scenario: Claude Code crashes (API error 500, timeout, killed process)
⚠️ Always check for stale processes first:
# Check if Claude is still running
ps aux | grep claude
# Kill any zombie processes for this project
pkill -f "claude.*projects/roundvision" || echo "No zombie processes"
# Also kill any hanging node processes
pkill -f "npx.*bmad" || echo "No zombie npx"
Check what was generated:
ls -lt _bmad-output/*.md | head -10
Find the last valid file:
# Read the most recently modified output
ls -t _bmad-output/*.md | head -1 | xargs head -30
Resume from where it stopped:
architecture.md exists but stories/ missing → run story generationstories/ exist but no OCM tasks → create tasks from existing storiesNever restart from zero if partial output exists
When Claude Code asks questions during execution:
process action:log sessionId:XXX to see what it askedprocess action:submitExample:
# Claude asks: "What's your preferred authentication provider?"
# If you don't know → ask user: "Claude needs to know auth provider - Auth0, Firebase, or Supabase?"
# Then provide the answer:
process action:submit sessionId:XXX data:"Auth0"
Rule of thumb: If it needs a story breakdown and sprint planning → BMad. If it's a simple edit → coding-agent directly.
After BMad workflows complete, documents are in:
project/
├── _bmad/
│ └── config.yaml
├── _bmad-output/
│ ├── brainstorming-report.md
│ ├── product-brief.md
│ ├── PRD.md
│ ├── ux-spec.md
│ ├── architecture.md
│ ├── epics/
│ │ └── epic-[name]/
│ │ └── stories/
│ │ └── story-[slug].md
│ └── sprint-status.yaml
└── project-context.md
⚠️ Always verify files exist by running ls _bmad-output/ or ls -R _bmad-output/ after each workflow.
Verify output validity before reading:
# Quick check
ls -la _bmad-output/architecture.md
# Validate content
head -20 _bmad-output/architecture.md
# Check for errors
grep -i "error\|fail" _bmad-output/architecture.md
⚠️ After Claude Code modifies source files, your cached view is stale!
Rule: After each successful Claude Code intervention on source code:
# Bad: Assuming old read is still valid
read path:"~/project/src/auth.js" # ❌ May be outdated
# Good: Read fresh after Claude modified it
exec command:"cat ~/project/src/auth.js" # ✅ Fresh content
Before moving to Implementation phase:
architecture.md (or tech-spec.md for quick-dev)| Error | Solution |
|---|---|
Command not found | Check PATH: echo $PATH and which claude |
npx: command not found | Install Node.js 20+ |
_bmad/ not found | Run npx bmad-method install first |
| Claude stuck on permission | Use --dangerously-skip-permissions |
| API 500 error | Trigger recovery (see "Recovery After Crash") |
| Session timeout | Check if process still running, resume if possible |
rm -rf via Claude Code without explicit human validationClaude Code often asks: "Do you want to commit these changes? [y/N]"
# When Claude asks to commit, default to "n"
process action:submit sessionId:XXX data:"n"
User: "On project PingRoot, analyze the current architecture and generate the product brief"
Agent does:
# 1. Pre-flight check
ls ~/projects/pingroot/_bmad/ || echo "Need to install BMad"
# 2. Check/update project-context.md
ls ~/projects/pingroot/project-context.md || echo "Need to create project-context.md"
# 3. Launch architecture workflow
bash pty:true workdir:~/projects/pingroot background:true command:"claude --dangerously-skip-permissions '/bmad-bmm-create-architecture'"
# 4. Monitor, wait for completion
process action:poll sessionId:XXX
# 5. Verify output
ls _bmad-output/architecture.md
head -20 _bmad-output/architecture.md
grep -i "error" _bmad-output/architecture.md || echo "OK"
# 6. If OK, verify coherence with user request
# 7. If coherent, launch product brief
bash pty:true workdir:~/projects/pingroot command:"claude --dangerously-skip-permissions '/bmad-bmm-create-product-brief'"
# 8. Deliver outputs
User: "Prepare sprint 1 for RoundVision and add tasks to OCM"
Agent does:
# 1. Pre-flight check
ls ~/projects/roundvision/_bmad/ || npx bmad-method install
# 2. Check project-context.md
ls ~/projects/roundvision/project-context.md || echo "Update this first!"
# 3. Launch sprint planning + story creation
bash pty:true workdir:~/projects/roundvision background:true command:"claude --dangerously-skip-permissions '/bmad-bmm-sprint-planning && /bmad-bmm-create-epics-and-stories'"
# 4. Monitor and wait for completion
process action:poll sessionId:XXX # repeat until done
# 5. Refresh context - verify files exist
ls -R ~/projects/roundvision/_bmad-output/epics/
# 6. List stories first (don't dump all at once!)
ls ~/projects/roundvision/_bmad-output/epics/*/stories/
# 7. Read and process stories one by one
for story in ~/projects/roundvision/_bmad-output/epics/*/stories/*.md; do
echo "=== $(basename $story) ==="
head -20 "$story"
# Create OCM task from this story
done
# 8. Report: "Created X tasks in OCM for Sprint 1"
# IMPORTANT: Each OCM task must include story path as reference!
User: "Fix the typo in the login page"
Agent does:
# Direct coding-agent, no BMad workflow needed
bash pty:true workdir:~/projects/login command:"claude 'Fix the typo on line 42: \"Passowrd\" → \"Password\"'"
Scenario: Claude Code crashes during story generation
Agent does:
# 0. Cleanup zombies FIRST!
ps aux | grep claude
pkill -f "claude.*projects/roundvision" || echo "Clean"
# 1. Check what was generated
ls -lt ~/project/_bmad-output/ | head -10
# 2. Find last valid file
ls -t ~/project/_bmad-output/epics/*/stories/ | head -1
# 3. Check if partial stories exist
ls ~/project/_bmad-output/epics/*/stories/*.md | wc -l
# 4. If partial → resume from last point
# If 3 stories out of 5 → generate remaining 2
# If 0 stories → restart story generation
# 5. Continue without restarting from zero
The minion does NOT automatically have access to project files! sub-agent to implement
When spawning a a task, you MUST provide:
# Minion needs workdir to access project files
sessions_spawn workdir:"~/projects/roundvision" ...
⚠️ NEVER give only the story to a minion!
The story says "Add a login button" but doesn't say:
You MUST provide:
# Step 1: Read all three
cat ~/projects/roundvision/project-context.md
cat ~/projects/roundvision/_bmad-output/architecture.md
cat ~/projects/roundvision/_bmad-output/epics/auth.md
# Step/stories/story-login 2: Combine into a comprehensive prompt
sessions_spawn task:"You are implementing this story: [STORY].
Project context: [CONTEXT].
Architecture: [ARCHITECTURE].
Follow the patterns defined in architecture.md."
{
"title": "Implement login form validation",
"description": "Full story content...",
"source": "_bmad-output/epics/auth/stories/story-login.md"
}
⚠️ Without workdir + story content + context + architecture, the minion is blind and cannot implement anything!
_bmad-output/project-context.md is the project's brain - keep it updated/bmad-help provides interactive guidancepty:true with Claude Code--model <sonnet|haiku|opus>