bdarbaz
bdarbaz/claude-stack-plugin
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/bdarbaz/claude-stack-plugin --skill s-help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。
Autonomous workflow loop - runs the full skill chain automatically with circuit breaker and stuck detection
Freeform router - accepts natural language and routes to the appropriate skill(s)
Generate structured codebase documentation - file tree, dependency graph, module responsibilities, entry points
基于 SOC 职业分类
正在显示 SKILL.md
| name | s-help |
| description | Show all available skills, agents, and workflow guides - quick reference card for claude-stack |
Display all available skills, agents, workflows, and usage guides. The one-stop reference card.
Show all 27 skills organized by phase group:
╔══════════════════════════════════════════════════════════════════╗
║ CLAUDE-STACK SKILLS (27) ║
╠══════════════════════════════════════════════════════════════════╣
║ ║
║ CORE (8) ║
║ ───────────────────────────────────────────────────────────── ║
║ /s:new Initialize project (.planning/ + docs/) ║
║ /s:discuss Gather requirements (REQUIREMENTS.md) ║
║ /s:plan Create roadmap and phase plans ║
║ /s:build TDD implementation (RED > GREEN > REFACTOR) ║
║ /s:status Show project state, progress, blockers ║
║ /s:resume Reconstruct context, continue work ║
║ /s:pause Save state, create resume instructions ║
║ /s:note Add quick note to STATE.md decisions ║
║ ║
║ QUALITY (6) ║
║ ───────────────────────────────────────────────────────────── ║
║ /s:brainstorm Explore ideas, propose approaches ║
║ /s:review Multi-perspective code review (5 aspects) ║
║ /s:verify Evidence-based verification (proof required) ║
║ /s:debug Systematic debugging (no guessing) ║
║ /s:ceo-review Product/business perspective review ║
║ /s:eng-review Technical/architecture perspective review ║
║ ║
║ SHIP (4) ║
║ ───────────────────────────────────────────────────────────── ║
║ /s:ship Test + commit + push + PR ║
║ /s:compound Save learnings to docs/solutions/ (NEVER SKIP) ║
║ /s:retro Retrospective: what worked, what didn't ║
║ /s:docs Update project documentation ║
║ ║
║ BROWSER (4) ║
║ ───────────────────────────────────────────────────────────── ║
║ /s:qa Browser QA: screenshot, test, fix (3 tiers) ║
║ /s:qa-only Browser QA: report only, no fixes ║
║ /s:browse Headless browser control (navigate, click, fill) ║
║ /s:cookies Import browser cookies for auth testing ║
║ ║
║ AUTO (5) ║
║ ───────────────────────────────────────────────────────────── ║
║ /s:auto Run full skill chain autonomously ║
║ /s:quick Compressed workflow for small tasks ║
║ /s:do Natural language router to skills ║
║ /s:map Generate codebase architecture docs ║
║ /s:help This reference card ║
║ ║
╚══════════════════════════════════════════════════════════════════╝
Show all 12 agents with their dispatch triggers:
╔══════════════════════════════════════════════════════════════════╗
║ CLAUDE-STACK AGENTS (12) ║
╠══════════════════════════════════════════════════════════════════╣
║ ║
║ Agent Dispatched By Expertise ║
║ ────────────────── ────────────────── ────────────────── ║
║ architect /s:plan, /s:brainstorm System design ║
║ code-reviewer /s:review 5-aspect review ║
║ security-reviewer /s:review OWASP, auth, XSS ║
║ database-reviewer /s:review Schema, queries, N+1 ║
║ tdd-guide /s:build RED/GREEN/REFACTOR ║
║ qa-engineer /s:qa Test matrix, edge ║
║ frontend-dev /s:build React, Next.js, CSS ║
║ backend-dev /s:build API, services, DB ║
║ mobile-dev /s:build React Native, Swift ║
║ debugger /s:debug Root cause analysis ║
║ performance-eng /s:review Profiling, optimize ║
║ deployment-eng /s:ship CI/CD, Docker, k8s ║
║ ║
║ + 4 subagent templates in agents/subagent-templates/ ║
║ implementer, spec-reviewer, code-quality, plan-reviewer ║
║ ║
╚══════════════════════════════════════════════════════════════════╝
Show the standard workflow progression:
STANDARD WORKFLOW CHAIN:
════════════════════════
/s:new ─→ /s:discuss ─→ /s:brainstorm ─→ /s:plan
│
┌────────────────────┘
▼
/s:build ─→ /s:review ─→ /s:verify
│ │
│ (TDD loop) │ (evidence check)
│ │
└──── fail? ────────────┘
│
▼ pass
/s:ship ─→ /s:compound
│
▼
/s:retro
SHORTCUTS:
/s:auto Runs the full chain automatically
/s:quick Compressed chain for small tasks
/s:do Natural language, routed to any skill
Show common starting scenarios:
QUICK START:
════════════
New project:
/s:new myproject
Start working on requirements:
/s:discuss
Resume from last session:
/s:resume
Check current status:
/s:status
Quick bug fix:
/s:quick "fix the login redirect issue"
Quick feature:
/s:quick "add dark mode toggle" --discuss
Full autonomous run:
/s:auto
Natural language (let the router decide):
/s:do "add user authentication with Google OAuth"
Map the codebase:
/s:map
Get a code review:
/s:review
Ship current work:
/s:ship
Show the 8 active rules:
ACTIVE RULES (8):
══════════════════
tdd.md No production code without a failing test
verification.md No "done" claims without evidence table
debugging.md No guessing - hypothesis > test > fix
context-persist Never delete .planning/ or docs/
ask-user.md Re-ground > Simplify > Recommend > Options
quality.md No 'any' types, no emoji icons, use Lucide/Heroicons
confidence.md Score confidence 1-10 before major decisions
uncodixfy.md 33 anti-AI-slop rules + 24 hard-no patterns
Show how project memory works:
CONTEXT SYSTEM:
═══════════════
.planning/ Project memory (created by /s:new)
STATE.md Current position, decisions, blockers
REQUIREMENTS.md ID'd requirements with acceptance criteria
ROADMAP.md Phases, tasks, dependencies
phases/N/ Per-phase context, plan, research, summary
docs/ Knowledge base (grows over time)
brainstorms/*.md Explored ideas and decisions
plans/*.md Implementation plans with checkboxes
solutions/*.md Past solutions and learnings
RULE: Never delete .planning/ or docs/ - they are project memory.
RULE: Every skill updates STATE.md.
RULE: /s:compound saves learnings after each feature.
Show useful tips for effective usage:
TIPS:
═════
1. Always start with /s:new for a new project
2. /s:compound after every feature - learnings compound over time
3. /s:status when you're lost - it reads STATE.md + git log
4. /s:resume at the start of each new session
5. /s:quick for small tasks - still enforces TDD
6. /s:do when you're not sure which skill to use
7. /s:debug instead of guessing at bug fixes
8. /s:map to understand an unfamiliar codebase
9. Agents are dispatched automatically by skills when needed
10. Split-pane teammates for parallel work (not subagents)