claude-code-session-transcript-analysis

You can recover decisions from a Claude Code session transcript by parsing the stored JSONL file to extract agent state, thinking blocks, and tool interactions. This skill detects interrupted sessions and aggregates the extracted context into a structured recovery document for continuation.

Auditing a long-running agent session requires systematically analyzing the JSONL transcript to reconstruct decisions, learnings, and final responses. This skill traces tool interactions and generates executive summaries alongside task status to provide a complete recovery framework.

Reconstructing agent state after hitting the context limit is done by parsing the Claude Code JSONL transcript to extract previous decisions and learnings. The skill aggregates this historical data into a structured recovery document, enabling seamless session continuation.

Yes, extracting thinking blocks and tool interactions from a JSONL session transcript is a core function of this analysis process. The skill parses the transcript file to detect these elements, allowing you to trace decisions and reconstruct the agent's internal state.

No, this session transcript analysis does not require external dependencies or libraries. You simply input the JSONL transcript file path to extract decisions, learnings, and agent state for recovery without needing additional environment setup.

Your Claude Code session recovery document may show incomplete task status if the JSONL transcript analysis was run on an interrupted session before it reached final responses. The tool extracts available decisions and tool interactions to reconstruct as much state as possible.