菜单
Use to run a full Raven boot diagnostic. Checks all files, agents, manifest, and hooks are correctly installed. If no manifest exists, runs first-run discovery instead of erroring.
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。
基于 SOC 职业分类
| name | raven-debug |
| description | Use to run a full Raven boot diagnostic. Checks all files, agents, manifest, and hooks are correctly installed. If no manifest exists, runs first-run discovery instead of erroring. |
Before running any diagnostics:
Is .raven/manifest.json present?
If .raven/manifest.json does NOT exist:
Do NOT error. Do NOT say "manifest missing — hard stop."
Instead, say this:
─────────────────────────────────────────────────
RAVEN — not set up yet for this project
─────────────────────────────────────────────────
No manifest found at .raven/manifest.json
Raven can help with this project. Here's what
I can see in the current directory:
[Run: python3 .claude/scripts/sr-detect-workmode.py .]
[Display the signals and detected mode from the output]
This looks like a [detected mode] project.
Want me to set Raven up now?
1) Yes — set it up (takes 2 minutes)
2) No — just tell me what Raven would do here
3) Show me the manifest that would be created
→
If user says 1: Run the setup flow inline (ask the 2-3 questions, create the manifest, confirm). If user says 2: Describe what Raven would enforce for the detected mode, no manifest created. If user says 3: Generate and display the manifest JSON without saving it. Ask "Save this? [y/n]".
This is the correct behaviour when a developer runs raven-debug as their first command — they're exploring. Reward that, don't punish it.
Run these checks in order. Output ✅ or ❌ per check.
.env, *.pem, *.key, manifest.secrets.json?─────────────────────────────────────────────────
RAVEN — diagnostic
─────────────────────────────────────────────────
Project: {project name from manifest}
Work type: {work_type}
Mode: {solo/team/enterprise}
Platform: {auto-detected}
✅ CLAUDE.md present
✅ manifest.json valid
✅ work_type: infra
⚠️ manifest.secrets.json missing — get from your architect
✅ .gitignore configured
✅ 5 agents loaded: manifest-checker stack-validator style-enforcer architecture-guard db-guard
✅ Skills: raven-core + 23 specialists
✅ pre-commit hook executable
✅ Hooks: PreToolUse PostToolUse PreCompact Notification
⚠️ .raven/architecture.md missing — create before first commit
CLEARED — 2 warnings
─────────────────────────────────────────────────
Final line:
CLEARED — all checks passed (warnings OK)CLEARED — N warning(s) — passed with warningsX ERROR(S) FOUND — run /raven-debug for details (only if hard errors)Initialize Raven for a new project. Andie-conversational — auto-scans project, asks at most 2 questions, generates manifest.json, validates, and commits with audit trail.
USE PROACTIVELY whenever the user asks for: planning, design, architecture decision, tradeoff analysis, comparing approaches, strategy, system design, refactor scope, deciding what to build, or any non-trivial request needing clarification. Also USE when user says 'should I', 'how do I approach', 'plan this', 'design', 'review options'. Compact plan-first orchestration. Routes work, runs triad (Functional/Technical/Data), HITL gated, OODA loop. Hands off plans, never implements. Brownfield bugs → andie-jr.
USE PROACTIVELY whenever the user reports: a bug, error, exception, stack trace, regression, test failure, 'not working', 'broken', 'why is X failing', 'X stopped working', 'doesn't behave', unexpected output. Brownfield debug assistant. Max 2 rounds, max 3 roles, returns problem · root cause · fix · why · audit note · commit suggestion. Do NOT use for greenfield design — that goes to andie.
USE PROACTIVELY whenever the user asks for: planning, design, architecture decision, tradeoff analysis, comparing approaches, strategy, system design, refactor scope, deciding what to build, or any non-trivial request needing clarification. Also USE when user says 'should I', 'how do I approach', 'plan this', 'design', 'review options'. Compact plan-first orchestration. Routes work, runs triad (Functional/Technical/Data), HITL gated, OODA loop. Hands off plans, never implements. Brownfield bugs → andie-jr.
Tool and product registry for Andie pre-flight. Maps problem types to tool options with tradeoffs. Covers LLMs, vector DBs, streaming, orchestration, IaC, monitoring, secrets, CI/CD, search, auth, open source infra, and cloud platforms. Load when a tool selection decision is in scope.
Tool and product registry for Andie pre-flight. Maps problem types to tool options with tradeoffs. Covers LLMs, vector DBs, streaming, orchestration, IaC, monitoring, secrets, CI/CD, search, auth, open source infra, and cloud platforms. Load when a tool selection decision is in scope.