ce-session-extract

To extract error signals from a Claude Code session JSONL file, pipe the file into the skill to surface failed tool results and error messages with timestamps. It outputs a deterministic, parse-friendly digest for quick triage.

You can extract a narrative skeleton from Codex or Cursor session logs by piping the JSONL file into the skill. It generates a compact narrative of user and assistant turns with collapsed tool calls to minimize context size.

The best way to reduce context size when analyzing session JSONL transcripts is to extract a filtered skeleton or error signals. This collapses tool calls and removes unnecessary content, producing a lean digest with a final _meta summary.

Session skeleton extraction works across Claude Code, Codex, and Cursor platforms for unified analysis. It supports parsing transcripts from these environments to keep context within limits during analysis.

To parse a single session JSONL file for a structured summary, pipe the file to the skill to receive a readable skeleton or error digest. The output includes a deterministic _meta summary containing line counts and parse results.