session-retro

Extract retrospective lessons from session transcripts into rule, sensor, issue, and eval-case proposals.

Updated May 17, 2022

Search Tags:#retrospective #transcript analysis #issue triage #guardrails #session workflow #evaluation cases

Ready to run

npx skills add https://github.com/kanade0404/dotfiles --skill session-retro-kanade0404

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: session-retro
Source: https://github.com/kanade0404/dotfiles/tree/main/.opencode/skills/session-retro
Command: npx skills add https://github.com/kanade0404/dotfiles --skill session-retro-kanade0404

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill turns the end of a work session into concrete follow-up proposals so repeated mistakes become process improvements instead of recurring friction.

Core Features & Use Cases

Signal-driven retrospection: Focuses only on high-value signals such as failed tool calls, human corrections, and escalations.
Action routing: Sorts insights into four outcomes: rule, sensor, issue, and eval-case.
Guardrail design: Recommends prevention rules and detection checks that can reduce repeat failures in future sessions.
Use Case: After finishing an implementation or release session, use this Skill to capture what should be documented, tested, deferred, or preserved as a reusable evaluation case.

Quick Start

このセッションの失敗シグナルを振り返り、rule、sensor、issue、eval-case に振り分けて提案してください。

Frequently Asked Questions about session-retro

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

FAQPage Schema

How do I extract lessons learned from a session transcript for issue triage?▼

To extract lessons learned from a session transcript, analyze high-value signals such as failed tool calls, human corrections, and escalations to identify specific friction points. This process transforms repeated mistakes into actionable process improvements.

What is the best way to categorize retrospective insights from a development session?▼

The best way to categorize retrospective insights is to sort them into four specific outcomes: rules, sensors, issues, and eval-cases. This action routing separates insights into prevention rules, detection checks, deferred problems, and reusable evaluation cases.

When should I run a retrospective analysis on my release or issue-driven work?▼

You should run a retrospective analysis at the end of shipping, release, or issue-driven work sessions. Applying it at these milestones captures what needs to be documented, tested, deferred, or preserved before context is lost.

How do I create guardrails to prevent repeated tool failures in future sessions?▼

To create guardrails against repeated tool failures, analyze the failure signals from your transcript and generate specific prevention rules and detection checks. These guardrails are categorized as rules and sensors to actively reduce repeat failures.

Can I generate reusable evaluation cases from session failures?▼

Yes, you can generate reusable evaluation cases from session failures by extracting the specific scenarios where tools failed or humans corrected the workflow. These events are routed into eval-case recommendations for future testing.

Does session retrospective analysis work without adding external dependencies?▼

Session retrospective analysis works without external dependencies by processing the transcript directly to identify failures and corrections. It relies solely on the session transcript content to generate follow-up proposals and guardrails.