What problem does it solve? Investigating what happened during an Intendant agent run requires navigating a complex log layout of session.jsonl timelines, turn files, byte spans, external-agent wrappers, and managed-context sidecars, which is slow and error-prone without a map. ## Core Features & Use Cases - Session Discovery: Find sessions by id, prefix, project root, or recency using session_meta.json and the external wrapper index. - Deep Log Search: Query session.jsonl events, transcript.jsonl, daemon.log, and turn files with jq and ripgrep recipes, including exact byte-span reads of model output and stdout/stderr. - Artifact Interpretation: Understand event taxonomy, context snapshots, file snapshot history, recordings, frames, Codex/Claude Code backend logs, and context rewind/fission sidecars. - Use Case: After an agent session fails, locate the session directory, inspect the event histogram, read the exact failing tool output span, and trace the error through daemon.log without reading the Intendant source code. ## Quick Start Ask the agent to use the intendant-log-search skill to find the most recent Intendant session for the current project and summarize what happened in it.
