What problem does it solve? Debugging the chess_teacher app requires knowing where logs live: an active local JSON-lines buffer, rotated .ready segments awaiting upload, and historical segments shipped to S3. This Skill gives a read-only, structured way to search, tail, and inventory logs across all three tiers without guessing paths or credentials. ## Core Features & Use Cases - Buffer status and tailing: Inspect active log files, pending upload counts, and the most recent JSON log lines from the local buffer. - Cross-tier search: Filter log records by level, logger, substring, environment, time window, date, or pod hostname across local files and/or S3 segments. - Segment inventory and reading: List local pending and S3 shipped segments, then read a single segment by relative path or S3 key. - Use Case: A user asks why ingestion failed at 3pm in production. The agent runs a search against S3 with --level ERROR and --since filters, summarizes the matching JSON records, and correlates with pod hostnames. ## Quick Start Ask the agent to search production logs for errors in the last 24 hours using the log query script with the prod Doppler config.
