What problem does it solve? Codebases often contain hidden reliability weaknesses—swallowed exceptions, network calls without timeouts, missing retries, non-idempotent writes, and absent health checks—that only surface as production outages or data loss. This Skill statically scans the repository for these design flaws and reports them by severity before they cause incidents. ## Core Features & Use Cases - Exception Handling Audit: Detects bare or over-broad exception handlers and logged-but-swallowed errors in Python, TypeScript, JavaScript, and Java code. - Fault Tolerance & Availability Checks: Finds external calls without timeouts, missing retry/backoff or circuit breakers, missing health/readiness endpoints, and absent graceful-shutdown handling. - Recoverability Analysis: Flags non-idempotent writes, multi-step writes without transaction boundaries, and background jobs without clear delivery semantics. - Use Case: Before scaling a service to multiple instances, run the audit to surface single-instance assumptions like hardcoded singletons and in-memory state that would break under horizontal scaling. ## Quick Start Ask the agent to run a reliability audit on the current repository, optionally scoped to a directory such as src/payments, and review the severity-classified findings report.
