When is the best time to run a session review?

You should run the session-review skill after completing a project epic, hitting a series of technical issues, or finishing a full development cycle to extract learnings.

Does it provide specific recommendations for other skills?

Absolutely. The output includes specific updates you can make to your other Claude Code skills, such as adding checklists or known-issue files to prevent repeated mistakes.

How does this skill help improve AI task estimation?

It compares the original time estimates in your tasks.db against the actual time logged, providing variance data and suggested multipliers for specific task types.

Can this skill identify why an agent is over-engineering a feature?

Yes, by analyzing session logs and git diffs against the original features.json, it can detect scope creep and suggest tighter checkpoint settings.

What data sources does the skill use for its analysis?

It analyzes the .claude/ directory (logs, progress.md, tasks.db) and your Git history (commit messages, diffs, and reverts) to get a full picture of the work.

Session Review

Name: Session Review
Author: srstomp

bysrstomp

•

생산성 및 워크플로

Analyzes completed development sessions to identify performance patterns, estimate accuracy, and actionable improvements for AI agents.

소개

The Session Review skill acts as a dedicated retrospective tool for Claude Code, meticulously examining session logs, git history, and task databases to evaluate agent performance. By comparing initial plans against actual execution, it uncovers behavioral patterns—such as recurring bugs or over-engineering tendencies—and provides structured reports with recommendations for skill updates and checkpoint adjustments. It is an essential tool for developers looking to move from supervised to autonomous AI workflows by learning from past sessions to optimize future performance.

주요 기능

Automated analysis of session logs, progress reports, and git history
Pattern detection for identifying recurring agent habits and errors
Generation of project artifacts like known-issues.md and estimate-adjustments.md
Actionable recommendation engine for prompt and skill improvements
2 GitHub stars
Quantitative comparison of planned vs. actual task completion and timing

사용 사례

Conducting a retrospective after completing a major project epic to improve future velocity
Calibrating task multipliers based on historical data for more accurate project planning
Troubleshooting agent 'thrashing' or repeated reverts in complex codebases

소개

주요 기능

Automated analysis of session logs, progress reports, and git history
Pattern detection for identifying recurring agent habits and errors
Generation of project artifacts like known-issues.md and estimate-adjustments.md
Actionable recommendation engine for prompt and skill improvements
2 GitHub stars
Quantitative comparison of planned vs. actual task completion and timing

사용 사례

Conducting a retrospective after completing a major project epic to improve future velocity
Calibrating task multipliers based on historical data for more accurate project planning
Troubleshooting agent 'thrashing' or repeated reverts in complex codebases