2026 07 12

---
title: netresearch/retro-skill
category: daily-digests/2026-07-12
tech_stack:
- Python
- Claude Code
- LLM (Claude)
- Bash
- Git
- GitHub CLI (gh)
- GitLab CLI (glab)
- jq
quality_score: 9
rag_relevance: 7
tags:
- session retrospection
- friction detection
- Claude Code plugin
- learning materialization
- agent session analysis
source: https://github.com/netresearch/retro-skill
discovered_at: '2026-07-12T10:26:19Z'
evaluated_by: mistral-small-latest
---

## Summary
retro is an LLM-driven plugin for Claude Code that analyzes agent session transcripts to detect friction, classify findings into six destinations, and materialize actionable learnings with explicit user approval. It replaces noisy background hooks with a focused, post-session review process.

## Key Features
- Post-session transcript analysis to detect friction (tool errors, repeated mistakes, convention violations)
- Six destination routes for findings: user-memory, project-rules, skill-update, new-skill, checkpoint, harness-artefact
- Explicit per-proposal approval for all materializations (no silent writes)
- Mechanical pre-pass + LLM enrichment pipeline for accurate classification and token efficiency
- Supports four modes: Sweep, Spotlight, Outcome, and Constitutional Audit for comprehensive session review

## Why It Matters for RAG Builders
It provides a structured way to extract and materialize actionable learnings from agent sessions, reducing noise and improving future agent performance by capturing both friction and reusable knowledge.

## Tech Stack Deep Dive
### Python
Automated review identified **Python** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### Claude Code
Automated review identified **Claude Code** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### LLM (Claude)
Automated review identified **LLM (Claude)** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### Bash
Automated review identified **Bash** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### Git
Automated review identified **Git** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### GitHub CLI (gh)
Automated review identified **GitHub CLI (gh)** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### GitLab CLI (glab)
Automated review identified **GitLab CLI (glab)** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.

### jq
Automated review identified **jq** as a key module contributing to infrastructure orchestration or cognitive reasoning boundaries in this project.



## Installation
```bash
# Install instruction (default)
pip install -r requirements.txt
```

## Related Vault Entries
<!-- Auto-populated by build-index.js based on tech_stack overlap -->