What problem does it solve?
Raw tool observation logs captured during AI coding sessions are unstructured and discarded after sessions end, leading to the permanent loss of valuable technical insights, decisions, and project context that could improve future work.
Core Features & Use Cases
- Processes JSONL observation buffers from Claude Code sessions to group related tool actions into coherent work units.
- Applies a strict quality rubric to filter out low-value noise, retaining only significant insights like technical decisions, bugfixes, recurring patterns, and project conventions.
- Submits approved memories directly to the Neuralscape memory system via MCP tools for persistent, searchable storage across sessions and projects.
- Use Case: If you debugged a dependency version conflict and pinned the correct library version during a session, this skill will capture that as a durable memory instead of letting the insight disappear when the session ends.
Quick Start
Invoke the compile-observations skill with the path to your session's JSONL observation buffer to automatically extract and save all significant memories from your recent tool usage.