skills.rest
fryanpan
fryanpanCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

retro

Run session retrospectives with transcript timing analysis and logged improvement actions.

Updated May 12, 2026
Search Tags:#claude code#retrospective#process improvement#transcript analysis#cost tracking#session timing#learnings log
One-click install
npx skills add https://github.com/fryanpan/ai-team-lead --skill retro-fryanpan
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: retro
Source: https://github.com/fryanpan/ai-team-lead/tree/main/plugin/team-lead-fleet/skills/retro
Command: npx skills add https://github.com/fryanpan/ai-team-lead --skill retro-fryanpan

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? After a long Claude Code session, it is hard to know where time actually went, what caused friction, and what should change next time. This Skill turns a finished session into a structured retrospective with measured timing data, concrete improvement actions, and a persistent log. ## Core Features & Use Cases - Transcript Time Analysis: Runs a deterministic Python script over the session JSONL transcript to compute hands-on time, agent time, per-turn costs, and error counts, rendered as a phase breakdown table. - Actionable Improvements: Converts observations and user feedback into specific deliverables such as skill edits, CLAUDE.md updates, doc changes, or tickets, with a parallel subagent auditing CLAUDE.md files. - Persistent Logging: Appends a formatted entry to docs/process/retrospective.md, elevates reusable insights into docs/process/learnings.md, and commits everything. - Use Case: After completing a multi-hour feature implementation, invoke /retro to see that 60 minutes went to hands-on review while the agent spent 30 minutes in fix cycles, then log a CLAUDE.md rule to prevent the repeated bug pattern. ## Quick Start Ask the agent to run /retro to analyze the current session transcript and log a retrospective with improvement actions.

Frequently Asked Questions about retro

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

FAQPage Schema
How do I run a retrospective on a Claude Code session?▼

Invoke /retro explicitly or trigger it after a full plan implementation completes. The skill runs scripts/analyze_transcript.py against the most recent session JSONL transcript, then walks through observations, feedback, actions, and logging.

How is hands-on time calculated from a Claude Code transcript?▼

The script estimates reading time at 150 words per minute of assistant output, typing time at 60 words per minute of user input, plus a one-minute buffer per turn. Consecutive turns with overlapping buffers merge into a single block to produce adjusted hands-on time.

Does the retro skill work for autonomous agent sessions?▼

Yes. In autonomous mode it skips gathering human feedback and uses only transcript observations. It executes low-risk reversible actions like doc updates and ticket creation, but defers skill or CLAUDE.md changes for human review.

What transcript files does the analysis script read?▼

It reads JSONL transcripts from ~/.claude/projects/<converted-cwd>/, selecting the most recently modified file. System-injected messages such as skill injections, command outputs, and system reminders are filtered out automatically.

Why does the retro skill not trigger automatically after every task?▼

It only runs on explicit invocation or after a complete plan implementation, by design. Suggesting a retro after every PR or review turns finishing work into a prompt, which interrupts flow, so the skill waits to be asked.

Related Skills

Explore More Software Engineering Skills→
qa-amanqa-aman
11

work-log

Generate daily work logs from Claude Code JSONL session transcripts.

Community
Intermediate
busboom-software-studiobusboom-software-studio
1

project-status

Analyze SE artifacts and tickets to generate a structured project status report.

Official
Intermediate
ivegamsftivegamsft

copilot-cli-usage-analytics

Analyze Copilot CLI session costs and model dispatches.

Community
Intermediate
robwestzrobwestz

heavy-file-ingestion-codex

Convert heavyweight PDFs, Office documents, and spreadsheets into markdown or CSV artifacts.

Community
Intermediate