session-retrospective

Captures session learnings across eight perspectives and routes them to knowledge destinations.

Updated Jan 30, 2026
One-click install
npx skills add https://github.com/RyoMa99/chezmoi_dotfiles --skill session-retrospective-ryoma99
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: session-retrospective
Source: https://github.com/RyoMa99/chezmoi_dotfiles/tree/main/dot_claude/skills/session-retrospective
Command: npx skills add https://github.com/RyoMa99/chezmoi_dotfiles --skill session-retrospective-ryoma99

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Development sessions generate valuable insights—domain knowledge, technical lessons, automation opportunities—that are usually lost when the session ends. This Skill structures the end-of-session retrospective so learnings are captured, reviewed, and stored in the right places instead of being forgotten. ## Core Features & Use Cases - Eight-Perspective Scan: Detects domain knowledge, technical learnings, automation opportunities, stale knowledge, CLAUDE.md gaps, estimate accuracy, RDRA artifact freshness, and design debt. - User-Approved Routing: Presents a structured report and saves only the items the user approves to destinations like ~/.claude/rules/, project CLAUDE.md, docs/rdra/overview.md, or GitHub Issues. - Automation Proposals: Classifies repeated patterns into hooks, skills, or rules with confidence scores and implementation drafts. - Use Case: After finishing a feature implementation, run the retrospective to discover that a recurring manual check should become a hook, a TypeScript error-resolution pattern belongs in ~/.claude/rules/typescript.md, and a design compromise should become a refactoring GitHub Issue. ## Quick Start Ask the AI to run the session retrospective and summarize this session's learnings across the eight perspectives for your approval.

Frequently Asked Questions about session-retrospective

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I capture learnings at the end of a Claude Code session?▼

Invoke the session-retrospective skill or ask the AI to run a retrospective. It scans the session across eight perspectives, presents a structured report of detected insights, and saves only the items you approve to the appropriate destinations.

What are the eight perspectives in a session retrospective?▼

The perspectives are domain knowledge, technical learnings, automation opportunities, stale knowledge updates, project CLAUDE.md gaps, estimate validity, RDRA artifact freshness, and design debt. Each has defined detection signals and a designated storage destination.

Where does the retrospective save technical learnings?▼

Technical learnings are routed to topic-specific rule files under ~/.claude/rules/, such as typescript.md, react-nextjs.md, or testing.md, based on the technology involved. Claude Code environment-specific workarounds go to troubleshooting.md.

Does the retrospective save findings automatically without approval?▼

No. The skill presents a report of detected insights and waits for explicit user selection (for example D1, T1, all, or none) before writing anything. Automatic saving is explicitly prohibited by its rules.

How are automation opportunities detected and classified?▼

Repeated manual checks, multi-step procedures, and recurring correction patterns are classified as hooks, skills, or rules with confidence scores from 0.3 to 0.9. Approved proposals include an implementation draft such as hook JSON or a SKILL.md template.

When should design debt become a GitHub Issue?▼

Design debt becomes a GitHub Issue when the gap between current and ideal design is large enough to affect future development speed. Minor gaps are only recorded in the report, and issues already resolved in the session are not recorded.