session-analyzer

You can validate completed Claude Code sessions by parsing debug and session JSONL logs to compare observed subagent invocations, hook matches, and artifact creation against a target SKILL.md, generating a structured PASS/FAIL discrepancy report.

Parsing Claude Code JSONL logs involves locating SubagentStart and SubagentStop events, hook permission decisions, and file write or delete operations with timestamps to verify actual execution behavior against expected workflow components.

Post-hoc session validation is the process of analyzing completed session logs to determine whether a skill followed its SKILL.md workflow, invoked expected subagents and hooks, and produced required artifacts, enabling engineers to trust past executions and diagnose failures.

Yes, you can verify artifact creation and hook permissions by parsing debug logs to extract file write and delete operations alongside hook match decisions, then comparing these observed events against the expected checklist derived from a target SKILL.md.

Generating a PASS/FAIL report requires extracting expected components from a target SKILL.md, parsing session logs to locate actual subagent and hook events, and producing a structured report that highlights deviations, ordering issues, missing artifacts, and hook failures.

Skill execution mismatches occur when observed subagent invocations, hook matches, or artifact creation events parsed from session logs deviate from the expected checklist extracted from the target SKILL.md, which the validation report highlights for debugging.