incident-readiness-check

name

incident-readiness-check

description

Walk a user through the shared-infrastructure incident-readiness check. Loads the responsibility-matrix and DPA definitions, gathers the org's assignments via dialogue, runs `siir check-responsibility` / `siir check-dpa` in JSON mode, and reports PASS/REVISE/BLOCK with the first gap to fix. Use when the user wants to evaluate whether a shared platform (OEM /共通基盤) is ready for the first 30 minutes of an incident.

incident-readiness-check

Interactively score a shared infrastructure's incident readiness against the definitions in this repository. This skill is a thin wrapper around the siir CLI: it gathers the responsibility-boundary assignments and DPA clause coverage via dialogue, runs the CLI in JSON mode, and translates the verdict back into a readable summary plus the first gap to fix.

When to use this skill

• The user runs a shared SaaS / OEM platform and wants to know whether the incident-initial-response responsibilities are clearly assigned
• The user asks to "do a readiness check" / "責任分界を点検" / "事故初動の備えを評価"
• The user wants to score against an overlay (their own company's stricter rules)

Workflow

1. Read definitions/responsibility-matrix.yaml to retrieve the 12 items and 4 roles. Do not hard-code them — always read the definition so overlays and version bumps stay in sync.

2. For each item, ask who is Accountable / Responsible / Consulted / Informed. Where the org genuinely has not decided, record tbd (the framework treats an explicit tbd as a healthy gray zone, not a failure).

3. Write the answers to a temp YAML:

   target: <platform name>
   matrix:
     RB01: { principal_isp: R, oem_operator: C, ops_bpo: I, sw_vendor: I }
     ...
   

4. Run bin/siir check-responsibility <tmp.yaml> --format json (add --overlay <path> for each overlay). Capture stdout and the exit code.

5. Optionally gather DPA clause coverage and run bin/siir check-dpa <dpa.yaml> --format json.

6. Translate the JSON: lead with the conclusion (PASS / REVISE / BLOCK), then list the items that are block (unassigned / no accountable / split A) and the items that are revise (gray zones to resolve). Recommend fixing the block items first, then re-running.

Failure modes to handle

• If bin/siir is not on PATH, fall back to python -m siir.cli ... with PYTHONPATH=<repo>/src.
• If the overlay fails check-overlay, surface the violation and stop rather than scoring with a half-applied overlay.