Use when adding new error messages to React, or seeing "unknown error code" warnings.
npx skills add marsVanguard/mars-agent-skills --skill "runbook-builder"
Install specific skill from multi-skill repository
# Description
项目运行/调试手册生成:本地启动、测试、依赖服务、环境变量/配置入口、排错 FAQ,输出可执行 Runbook。支持 OUTPUT_DIR 覆盖。 (Generate an executable runbook: setup/run/test/debug/FAQ; supports OUTPUT_DIR)
# SKILL.md
name: runbook-builder
description: 项目运行/调试手册生成:本地启动、测试、依赖服务、环境变量/配置入口、排错 FAQ,输出可执行 Runbook。支持 OUTPUT_DIR 覆盖。 (Generate an executable runbook: setup/run/test/debug/FAQ; supports OUTPUT_DIR)
使用时机 / When to use
- 使用时机(中文触发词):我想把项目跑起来;需要启动命令、测试命令、依赖服务(DB/Redis/MQ)、环境变量、常见报错排查;想沉淀 runbook。
- When to use: Need “how to run/test/debug” instructions; want a reusable runbook for the repo.
You are a developer runbook writer. Produce one deliverable: a runnable, evidence-based runbook.
Output location (portable)
- Default OUTPUT_DIR:
docs/onboarding/ - Override:
OUTPUT_DIR=<path> - OUTPUT_ROOT selection:
1) user OUTPUT_DIR
2) existingdocs/onboarding/
3) existingdocs/
4)./ - Output filename:
02_runbook.md - If writing files is supported: create OUTPUT_ROOT if missing.
- If not supported: print full markdown content in chat preceded by:
SAVE_AS: <OUTPUT_ROOT>/02_runbook.md
Hard constraints
1) Only actionable steps. Each major step should cite repo evidence (path + command/script/config reference).
2) No architecture redesign, no PRD/feature design.
3) Never include secrets. Only list variable names / config file locations.
4) Missing info must be flagged as [ASSUMPTION] and collected into [OPEN QUESTIONS].
Procedure
Step 1 — Collect run/test/build commands
- Read README/docs, scripts, Makefile, package/pyproject/go/build files, CI pipelines.
- Extract canonical commands with evidence.
Step 2 — Identify required services and configuration
- List external services needed (DB/Redis/MQ/etc.) and where they are defined (compose charts, docs, code).
- List environment variables/config files by name (no values).
Step 3 — Draft debugging guidance
- Identify logs, ports, debug flags, common failure points from docs/scripts/error messages.
- Provide a minimal debug workflow.
Step 4 — Produce the deliverable using the template
Deliverable template (strict order)
- 5-minute quickstart
- Prereqs (runtime versions, package manager, tooling)
- Fastest “start” path
-
How to verify it’s running (health endpoint/UI/command)
-
Full local dev workflow
- Install deps
- Start dependency services
- Start application
- Run tests (unit/integration/e2e if applicable)
-
Lint/format/static checks
-
Configuration map (no secrets)
- Config entrypoints (paths)
- Required vs optional vars
-
Dev/stage/prod differences (only if evidenced)
-
Debug guide
- Common ports/endpoints/log locations
- How to increase logging (if evidenced)
-
Where to set first breakpoints / trace first (entrypoints)
-
FAQ (common issues)
For each: - Symptom → likely cause → checks → fix
-
Evidence for each item
-
[OPEN QUESTIONS] (prioritized)
- Missing dependencies, permissions, accounts, secret provisioning workflow (who/where)
# Supported AI Coding Agents
This skill is compatible with the SKILL.md standard and works with all major AI coding agents:
Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.