gh-session-extract

To extract a conversation skeleton from a JSONL session file, provide the absolute file path and select skeleton mode. The skill parses the JSONL data to generate a compact narrative digest of user messages, assistant replies, and collapsed tool-call summaries for quick review.

Surfacing error signals from large session transcripts is done by running the skill in errors mode. This extracts tool failures and error events from the provided JSONL file, allowing quick triage without loading the massive original transcript into context.

Yes, you can preview a JSONL session file without loading the entire transcript by generating a readable digest. The skill creates compact narrative summaries of pre-selected sessions, enabling targeted session research while avoiding the context overhead of massive files.

Session extraction works without external dependencies, relying solely on its internal scripts. You only need to provide the absolute path to the target JSONL session file to parse the data and emit the narrative digests.

Use skeleton mode when you need a compact narrative of the overall conversation flow, including user and assistant messages. Use errors mode when your goal is specifically to surface and triage tool failures and error signals within the session file.