What did my agents run today? A daily note for a friend who lives in the terminal, written by Gemma 4 on his own laptop

Description and use

whatran automatically processes coding agents’ command history, flags risky operations, and generates a concise, model‑verified summary so developers can quickly see what agents did each day.

Problem

Coding agents execute shell commands that are recorded but not easily understood; users need a concise daily summary of what agents did, where, and any risky actions.

Outcome

whatran provides a daily, verified summary of coding agents’ terminal activity, flagging risky commands and explaining them in plain language, improving transparency and security for developers who rely on agents.

Workflow

• 1. Install atuin hook for Claude Code: atuin hook install claude-code. 2. Run agents; their commands are recorded in atuin and Claude Code session files. 3. Execute whatran --db demo/history.db --since 2d to generate the daily report. 4. whatran parses the SQLite database and session files, applies regex rules to flag risky commands. 5. The tool counts commands per agent/project, identifies failures, and extracts intent from session descriptions. 6. Gemma 4 model (served by Ollama) receives JSON facts and writes a concise paragraph per flagged event. 7. The note is cross‑checked against facts; if any mismatch, the note is suppressed and only counts/flags are shown. 8. The final report is displayed to the user, highlighting where agents worked, what deserves a second look, and a plain‑English explanation of each flagged action.

Evidence

whatran reads that history once a day and answers three questions: Where did the agents work? What deserves a second look? What was the agent trying to do?

Source

The commands those agents run are recorded, and nobody reads them. atuin can now capture what an agent runs (atuin hook install claude-code), tagged with the agent's name, and then it hides those entries from search by default so they don't clutter your own history.

Source

Gemma 4 wrote the note in about 11 seconds on my laptop: ... In plain words (gemma4:12b, running on this machine) ...

Source

whatran is Python with no dependencies of its own: sqlite3 and urllib from the standard library.

Source