What problem does it solve? Slow or error-prone API endpoints often go unnoticed until users complain, and manually digging through Datadog traces to find and explain bottlenecks is repetitive work. This Skill automates the sweep: it discovers which APIs were hit in a time window, flags the ones breaching latency or error thresholds, explains why they are slow by cross-referencing traces with the repository's code, and keeps one living report file per API per environment. ## Core Features & Use Cases - Automated traffic discovery and gating: Aggregates Datadog spans by resource name, computes p50/p95/error rates, and flags APIs exceeding configurable thresholds (default p95 > 200ms or error rate ≥ 5%). - Trace-to-code root cause analysis: Samples representative traces per flagged API, builds a time breakdown of internal calls, and maps each bottleneck to a concrete code site using a heuristics catalog (cache bypass, N+1, sequential calls, and more). - Idempotent living reports: Maintains fingerprinted per-API Markdown reports with a four-status taxonomy and append-only History, rewriting files only when findings genuinely change, plus a worst-first per-environment index. - Use Case: Schedule a daily run against production to continuously track slow endpoints; each morning the team gets updated report files committed and pushed to the findings repository, with History rows explaining what improved or regressed. ## Quick Start Ask the assistant to run an API performance scan of this service and update the performance reports for the last day of production traffic.
