What it does

Mines AI session logs for patterns that reveal what you actually spend time on, which tools get used most, and where repetitive work hides. Turns raw JSONL session data into actionable analysis.

The pattern

AI coding assistants log every tool call, file read, command execution, and skill invocation as structured JSONL. This data is a goldmine sitting in plain sight.

Parse session logs to extract:

• Tool usage frequency. Which tools dominate? If 40% of calls are file reads and 30% are grep, you are doing a lot of exploration. That might mean your project structure needs better conventions or your documentation is stale.
• Skill invocations. Which skills actually get triggered? Which ones never fire? Dead skills are noise in your context.
• Recurring task shapes. If you run the same sequence of commands across sessions (read config, check status, update file, commit), that sequence should be a script or skill.
• Time-per-task estimates. Measure elapsed time between the first and last tool call in a task cluster. Spot tasks that take 10x longer than expected.

The analysis output is a markdown report with three sections: statistics (raw numbers and distributions), patterns (repeated sequences and anomalies), and suggestions (concrete actions like "automate this 4-step sequence" or "retire this unused skill").

Key decisions

Parse locally, not with an external service. Session logs may contain sensitive context. Process them on your machine with simple Python or jq scripts.

Focus on sequences, not individual calls. A single file read tells you nothing. The pattern of read-grep-edit-test repeated 15 times tells you that workflow needs a tighter feedback loop.

Compare across time windows. One session is anecdotal. A week of sessions reveals habits. Diff weekly summaries to track whether automation investments actually reduce manual work.

Output suggestions, not just data. Raw statistics require interpretation. The script should identify the top 3 candidates for automation and explain why.

When to use it

Run session insight analysis weekly or after intensive project sprints. It is most valuable when you feel busy but cannot articulate where time goes. The gap between perceived and actual time allocation is usually large. Use the findings to prioritize which workflows to automate next.