Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate operational runbooks from codebase analysis covering deployment, incident response, scaling, and monitoring, with copy-paste commands and rollback steps. Use when bootstrapping ops docs, preparing for on-call, or post-incident.
.claude/skills/borghei-runbook-generator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 68% | 0% |
Analyze a codebase and generate production-grade operational runbooks with copy-paste commands, verification checks after every step, rollback procedures for every destructive action, escalation paths with contact information, and time estimates for capacity planning. Detects the stack (CI/CD, database, hosting, containers) and produces runbooks tailored to the actual infrastructure. Includes staleness detection to flag runbooks when referenced config files change.
Keywords: runbook, operational procedures, incident response, deployment, rollback, database maintenance, scaling, monitoring, on-call, SRE, postmortem
Before generating the runbook, confirm these inputs. If any is unknown or vague, ASK — do not assume:
Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
| Tool | Purpose | Command | |------|---------|---------| | runbook_scaffolder.py | Generate runbook markdown templates from a JSON service definition | python scripts/runbook_scaffolder.py -i service.json --type deployment -o runbook.md | | runbook_validator.py | Validate runbook markdown for completeness and quality (required sections, VERIFY blocks, hardcoded creds, escalation table) | python scripts/runbook_validator.py --dir docs/runbooks --strict | | staleness_checker.py | Check runbook freshness against configurable staleness thresholds | python scripts/staleness_checker.py docs/runbooks --threshold 90 --json |
Load the reference that matches the task — keep this file lean and pull detail on demand:
This skill covers:
This skill does NOT cover:
ci-cd-pipeline-builder for automated pipelines)migration-architect for schema migration tooling)observability-designer for monitoring infrastructure)skill-security-auditor for security-focused analysis)| Skill | Integration | Data Flow | |-------|-------------|-----------| | ci-cd-pipeline-builder | Runbook deployment steps align with pipeline stages | Pipeline config feeds into deployment runbook generation; runbook rollback steps reference pipeline rollback triggers | | observability-designer | Monitoring runbook references alert rules and dashboards | Observability outputs (alert names, dashboard URLs) are embedded in runbook VERIFY and Monitor steps | | migration-architect | Database maintenance runbook uses migration tooling conventions | Migration file paths and commands flow into the database runbook template; rollback steps mirror migration rollback commands | | release-manager | Release process triggers runbook execution checkpoints | Release tags and changelogs feed into runbook staleness checks; release gates reference runbook pre-deployment checklists | | env-secrets-manager | Runbook commands reference env vars managed by secrets tooling | Secret names and vault paths flow into runbook env var references; rotation schedules inform runbook update cadence | | changelog-generator | Post-deployment runbook steps cross-reference changelog entries | Changelog diffs help identify which runbook steps need re-verification after a release |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,008 | 6,014 | +20% | 1 | 1 | 0% | 224 | 1,772 | +691% | 0 | 0 | — |
case-02 | fail→pass | 33,295 | 6,222 | -81% | 1 | 1 | 0% | 5,748 | 2,350 | -59% | 0 | 0 | — |
case-03 | fail→fail | 25,100 | 9,369 | -63% | 1 | 1 | 0% | 4,373 | 2,089 | -52% | 0 | 0 | — |
case-04 | fail→fail | 11,181 | 18,129 | +62% | 1 | 1 | 0% | 2,137 | 4,337 | +103% | 0 | 0 | — |
case-05 | fail→fail | 17,667 | 19,183 | +9% | 1 | 1 | 0% | 3,277 | 5,272 | +61% | 0 | 0 | — |
case-06 | fail→fail | 19,723 | 25,966 | +32% | 1 | 1 | 0% | 3,880 | 6,364 | +64% | 0 | 0 | — |
case-07 | pass→pass | 14,281 | 15,199 | +6% | 1 | 1 | 0% | 908 | 2,450 | +170% | 0 | 0 | — |
case-08 | fail→fail | 17,466 | 8,620 | -51% | 1 | 1 | 0% | 3,194 | 2,754 | -14% | 0 | 0 | — |
case-09 | fail→pass | 15,727 | 7,410 | -53% | 1 | 1 | 0% | 2,427 | 1,955 | -19% | 0 | 0 | — |
case-10 | fail→pass | 44,514 | 21,196 | -52% | 1 | 1 | 0% | 5,455 | 4,129 | -24% | 0 | 0 | — |
case-11 | fail→pass | 17,209 | 21,317 | +24% | 1 | 1 | 0% | 2,770 | 4,677 | +69% | 0 | 0 | — |
case-12 | fail→pass | 15,988 | 16,808 | +5% | 1 | 1 | 0% | 2,363 | 3,980 | +68% | 0 | 0 | — |
case-13 | pass→pass | 4,300 | 7,776 | +81% | 1 | 1 | 0% | 615 | 2,518 | +309% | 0 | 0 | — |
case-14 | fail→pass | 6,034 | 10,911 | +81% | 1 | 1 | 0% | 981 | 3,057 | +212% | 0 | 0 | — |
case-15 | pass→pass | 11,131 | 6,526 | -41% | 1 | 1 | 0% | 1,644 | 2,289 | +39% | 0 | 0 | — |
case-16 | pass→pass | 8,605 | 7,416 | -14% | 1 | 1 | 0% | 1,269 | 2,509 | +98% | 0 | 0 | — |
case-17 | pass→pass | 10,952 | 6,592 | -40% | 1 | 1 | 0% | 1,539 | 2,333 | +52% | 0 | 0 | — |
case-18 | fail→pass | 21,658 | 27,623 | +28% | 1 | 1 | 0% | 3,387 | 5,943 | +75% | 0 | 0 | — |
case-19 | pass→pass | 15,527 | 14,271 | -8% | 1 | 1 | 0% | 2,367 | 3,501 | +48% | 0 | 0 | — |
case-20 | pass→pass | 5,409 | 4,154 | -23% | 1 | 1 | 0% | 725 | 1,961 | +170% | 0 | 0 | — |
case-21 | fail→pass | 9,204 | 2,861 | -69% | 1 | 1 | 0% | 1,418 | 1,784 | +26% | 0 | 0 | — |
case-22 | pass→pass | 8,703 | 9,437 | +8% | 1 | 1 | 0% | 1,447 | 2,854 | +97% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.