introspect

Analyze Claude Code session logs - extract thinking blocks, tool usage stats, error patterns, debug trajectories. Triggers on: introspect, session logs, trajectory, analyze sessions, what went wrong, tool usage, thinking blocks, session history, my reasoning, past sessions, what did I do.

1.08x

Quality

96%

Does it follow best practices?

Impact

90%

1.08x

Average score across 3 eval scenarios

Securityby

Risky

Do not use without reviewing

Evaluation results

74%

-4%

Session Intelligence Report

Tool usage statistics and thinking block extraction

Criteria

Without context

With context

Tool frequency extraction

100%

83%

Tool sort order

100%

75%

Thinking from content array

100%

Thinking content present

100%

Session duration

75%

100%

Summary extraction

100%

Token estimation

100%

Correct tool count

33%

Output file produced

100%

Subagent exclusion

16%

Without context: $0.2160 · 1m 7s · 10 turns · 15 in / 4,302 out tokens

With context: $0.6831 · 2m 20s · 28 turns · 278 in / 8,682 out tokens

96%

Database Migration Failure Investigation

Error pattern analysis from session logs

Criteria

Without context

With context

Error source: user entries

100%

Error source: tool_result

100%

Error regex pattern

80%

Number stripping

80%

Path stripping

90%

100%

Correct error count

100%

Cross-session analysis

100%

Error frequency ranking

100%

Tool ID linkage

100%

Output file produced

100%

Without context: $0.3032 · 1m 24s · 11 turns · 16 in / 5,853 out tokens

With context: $0.4724 · 1m 40s · 16 turns · 19 in / 6,963 out tokens

100%

22%

Developer Productivity Report: Session Analysis

Session export formats and file modification tracking

Criteria

Without context

With context

JSON has session_id field

100%

JSON has entries count

100%

JSON tools array structure

20%

100%

JSON summaries array

25%

100%

Markdown has Summary section

100%

Markdown Tool Usage table

100%

Files edited tracking

100%

Files written tracking

100%

Correct tool counts

100%

Both sessions exported

100%

Subagent files excluded

100%

Without context: $0.5521 · 2m 16s · 20 turns · 24 in / 9,325 out tokens

With context: $0.8981 · 3m 8s · 28 turns · 33 in / 12,931 out tokens

Repository: NeverSight/skills_feed
Commit: 5342bca

Evaluated: 13 days ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Session Intelligence Report Database Migration Failure Investigation Developer Productivity Report: Session Analysis

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.