Session Log Analyst

The problem

Finding specific information buried in months of agent logs is painful. Manually grepping through thousands of JSONL entries with inconsistent schemas leads to incomplete results and wasted time.

What it does

• Identifies log roots and probes varied record schemas across different agent frameworks.
• Indexes conversation history by timestamp and session size for targeted retrieval.
• Extracts specific user or assistant turns while filtering out noise and tool calls.
• Performs keyword searches across entire corpora to locate specific decisions or references.
• Aggregates usage statistics including message counts and costs grouped by date.

Frameworks & tools

Works with jq, find, and xargs. Compatible with Hermes, OpenClaw, and Claude projects log structures.

Why this beats prompting it yourself

Writing complex, schema-aware jq filters on the fly is error-prone. This skill handles schema variations like nested message objects and diverse timestamp formats, ensuring you get a clean table instead of a wall of JSON.

Use cases

• Recall specific technical decisions or promises made in conversations from last month.
• Generate a summary of topics discussed across a specific workspace over a date range.
• Audit LLM costs and message volume per day for a particular agent or project.
• Locate every session where a specific database or framework was mentioned.

Known limitations

Requires logs to be in JSONL format. Compressed archives must be decompressed before analysis, and extremely large corpora should be sampled first.