CatalogueDelivery & Operations
Resilience & Failure-Mode Audit
Asks what happens when a dependency, a zone, or a process stops, and what the system does not survive.
What it examines
- Calls with no timeout
- Retries that can amplify a failure
- Single components whose loss stops the product
- Recovery that depends on a person noticing
What you get back
A findings file. Each item has the evidence, how serious it is, and a fix.
Dependencies
- An AI coding agent that reads SKILL.md files, such as Claude Code, Codex or Gemini CLI
- Read access to the repository: outbound HTTP clients, database and broker clients, cloud SDKs, gRPC and internal service calls
- Python 3.11 or newer with the jsonschema package, to check findings.json before fixing anything
- Optional: infrastructure and deployment topology, including any service mesh or gateway that imposes timeouts or retries outside application code
- Optional: latency percentiles from telemetry. Without them, suggested timeouts are marked provisional.
- Optional: backup and disaster recovery documents. Without them the recovery ladder is insufficient evidence, not an assumed state.
- Optional: a trustworthy E4 finding set. The remediate skill needs it before it goes past a plan.
- Optional: a non-production environment, if you want fault-injection checks after remediation
How it works
The audit is read-only. It runs six checks: outbound timeouts, retry and backoff safety, circuit breakers, single points of failure, the backup and disaster recovery ladder, and partial failure handling at each integration boundary.
It never injects faults into production or restores over live data. Recommendations are ordered by blast radius, and missing timeouts usually come first.
The remediate skill reads findings.json only. Timeouts come from observed p99 where available, backoff with jitter is added only to idempotent operations, and every circuit breaker gets a fallback you agree on. Verification is by fault injection in non-production only. Backup and recovery work is a plan and checklist, never a live restore or drill.
Setup
- Unzip the download, then copy the audit and remediate folders into your agent's skills folder (for Claude Code, .claude/skills/) in the repository you want to check.
- State your availability commitment, or note that you have none.
- Ask your agent: Run a resilience and failure-modes audit on [scope]. Availability commitment: [state it, or none]. Mode: audit only.
- Check the result with the validator: pip install jsonschema, then python3 schema/validate.py findings.json.
- Read findings.md. Once you have reviewed the findings, run the remediate skill using findings.json, or re-run with Mode: audit and remediate.
Run it with your agent
Copy the prompt, fill in the parts in square brackets, and paste it into your agent.
Run the e5-resilience-failure-mode-audit skill on this repository. Availability commitment: [state it, or none]. Never inject faults or restore anything. Mode: audit only. Do not change any code. Write findings.json and findings.md in the repository root.
Run the e5-resilience-failure-mode-remediate skill using findings.json. Work through the findings I choose, and leave the rest open.