mickolasjae

anthropic-api-resilience

Use when adding rate-limit / overload resilience to any Anthropic SDK client (Claude API consumer). Generates round-robin key rotation + exponential backoff (1s→5min, 2x multiplier) + circuit breaker (10 consecutive failures). Triggers on: 'Anthropic API rate limit', 'Claude API retries', 'multiple Anthropic keys', 'round-robin Anthropic', 'production-grade Anthropic client', 'rate limit error 429', 'Anthropic overloaded_error'.

evidence-gated-ci

Use when adding CI gates to a new project that touches production state or external services. Scaffolds smoketest + e2e + artifact-verified-mutation discipline so no claim of state change ships without an HTTP response body / bridge ack on disk. Triggers on: 'add CI to this project', 'evidence-gated CI', 'pre-push gate', 'smoketest setup', 'how do I make sure my agent isn't lying about external mutations', 'artifact-verified'.

mcp-directory-readiness

Use when preparing an MCP server for submission to the Anthropic MCP Directory. Audits tool annotations, OAuth/HTTPS posture, and generates a gap list + pre-filled submission form data. Triggers on: 'submit to Anthropic Directory', 'MCP submission', 'audit my MCP server', 'is my MCP ready for the directory', 'directory submission checklist'.

multi-model-routing

Use when wiring multiple LLM providers / models into one application and you want to pick the cheapest model per task type without hand-coding the routing. Generates a router with per-task-type model selection, per-call cost telemetry, and a dashboard-ready emit. Triggers on: 'route between Claude and GPT', 'cheapest model for X', 'multi-model dispatch', 'LLM cost telemetry', 'which model should I use for vocab extraction', 'reduce LLM bill'.