Skip to content
Back to skills

Create Runbook

ASecurity

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.

  • 666 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added February 8, 2026
documentationpythongo

Security analysis

A100/100

Scanned February 12, 2026

npx -y skills add majiayu000/claude-skill-registry --skill create-runbook --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Create Runbook?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Create Runbook
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majiayu000-create-runbook/badge)](https://www.skillsdirectory.com/skills/majiayu000-create-runbook)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: create-runbook
description: 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.
user-invocable: true
---

# 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:

```markdown
---
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]
   ```

2. **[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

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.