What problem does it solve?
Long-running chat timelines accumulate noise such as reload notices, compaction messages, greetings, and one-word acknowledgments, bloating the SQLite message database and making history hard to browse. This Skill removes those low-value messages in bulk while protecting messages that carry user-uploaded media.
Core Features & Use Cases
- Pattern-based deletion: Match messages by SQL LIKE patterns or exact content, filtered by sender, length, rowid, or timestamp, with dry-run preview before deleting.
- Built-in cleanup groups: Run
cleanupAll() to sweep twelve predefined groups covering reloads, compaction, greetings, slash commands, git operations, build noise, and package installs.
- Session scope resolution: Automatically expands
web:* chats to the full session tree so branch chats are cleaned alongside the root chat.
- Optional vacuum: Copies the database to /tmp, vacuums it, and swaps it back with a brief piclaw restart to reclaim disk space.
- Use Case: After weeks of agent sessions, run a dry-run cleanup with the built-in groups to preview matches, then delete thousands of noise messages and vacuum the database to shrink it.
Quick Start
Ask the agent to run the timeline cleanup in dry-run mode with the built-in pattern groups to preview which low-value messages would be deleted from the web chat history.