OpenParable
Skip to content
← Back to skills

Create Runbook

Productivity
create-runbook

Create Runbook is an Agent Skill for productivity published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Extract a reusable runbook from a successful investigation or troubleshooting session. Use when Claude has just completed a multi-step diagnosis that could help future investigations.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/data/create-runbook/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/create-runbook, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Create Runbook

Extract procedural knowledge from this conversation into a reusable runbook.

When to use this

After completing a multi-step investigation that:

  • Spanned multiple datasources (Xatu, Prometheus, Loki, Dora)
  • Followed a diagnostic pattern others could reuse
  • Discovered insights that would help future investigations

Runbook format

Create a markdown file with YAML frontmatter in the runbooks/ directory:

---
name: [Imperative title, e.g., "Investigate Finality Delay"]
description: [1-2 sentence summary for semantic search matching]
tags: [keywords for search, 3-6 tags]
prerequisites: [datasources needed, e.g., xatu, prometheus, dora]
---

[Opening paragraph explaining WHEN this runbook applies and WHAT problem it solves.
Use MUST/SHOULD/MAY keywords inline to indicate requirement levels.]

## Approach

1. **[Step title]** - [Description with MUST/SHOULD/MAY constraints inline]

   ```python
   [Optional: Example code if helpful, but prefer referencing search_examples]
  1. [Next step] - Use search_examples("relevant query") for the query pattern. You SHOULD [constraint]. You MAY [optional action].

Key Thresholds

[If applicable, include a table of healthy/warning/critical thresholds]

Notes

  • [Key insight or gotcha learned during the investigation]
  • [Threshold values or timing considerations]

## Constraint keywords (RFC 2119)

Use these keywords inline in the prose to indicate requirement levels:

- **MUST** - Non-negotiable requirement. The investigation will fail or produce wrong results without this.
- **SHOULD** - Strongly recommended. Skip only with good reason and document why.
- **MAY** - Optional. Use judgment based on context and time available.

## Guidelines for extraction

1. **Focus on the diagnostic pattern**, not the specific incident details
2. **Reference examples** instead of embedding queries - use `search_examples("...")`
3. **Include key thresholds** discovered during the investigation
4. **Document gotchas** - what would have tripped you up without prior knowledge?
5. **Keep it actionable** - every step should tell the user what to do, not just what to think about

## Naming conventions

- File name: `kebab-case.md` (e.g., `finality_delay.md`, `block-propagation.md`)
- Runbook name: Imperative mood (e.g., "Investigate X", "Diagnose Y", "Debug Z")
- Tags: lowercase, single words or short phrases

## Output

Save the runbook to: `runbooks/[kebab-case-name].md`

After creating the runbook, verify it loads correctly:
1. The server should log "Runbook registry loaded" with the updated count
2. The runbook should be searchable via `search_runbooks`

$ARGUMENTS