Agent Retro

Run a retrospective on a coding-agent session and improve the agent's environment, not the code. Combines Derby & Larsen's five-stage retrospective, Shigeo Shingo's poka-yoke (turn a mechanical mistake into a check that makes it impossible), and Matt Pocock's retro categories — navigation, automated checks, coding standards, steering files, tool economy, information access. Covers mining the session transcript for corrections, wrong turns and slow searches; deciding where each fix belongs (analyzer, hook, CI, reviewer standards, CLAUDE.md pointer, skill, doc or memory); the no-op test for stee

5 stars 2 forksAI, LLMs & Agents Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add vnovakovits/claude-skills

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add vnovakovits/claude-skills --skill agent-retro
Manual install
git clone --depth 1 https://github.com/vnovakovits/claude-skills.git /tmp/claude-skills
mkdir -p ~/.claude/skills
cp -r /tmp/claude-skills/plugins/engineering-practices/skills/agent-retro ~/.claude/skills/agent-retro

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Agent Retro

Apply this skill after a coding-agent session, especially one that needed many corrections, took a long detour, or shipped a mistake that review caught. The question is not *"what was wrong with the code?"* but *"what about the environment let the agent go wrong, and what is the cheapest durable change that stops it next time?"* The environment is everything the agent works inside: steering files (`CLAUDE.md`, `AGENTS.md`), skills, docs, linters and analyzers, hooks, CI, tool access and memory.

It combines three sources. Esther Derby and Diana Larsen's *Agile Retrospectives* gives the structure. Shigeo Shingo's **poka-yoke** (mistake-proofing) gives the central rule: don't remind people not to make a mechanical mistake, make the mistake impossible or immediately visible. Matt Pocock's `retro` skill gives the categories of agent environment to inspect.

## Core Philosophy

**Fix the environment, not the agent.** Kerth's Prime Directive, applied to an agent: assume it did the best it could with the context, tools and instructions it had. Blame produces louder instructions; analysis produces better context.
Read the full skill on GitHub

Details

Skill name
agent-retro
Path
plugins/engineering-practices/skills/agent-retro/SKILL.md
License
MIT
Repo updated
16 November 2026

FAQ

What does the Agent Retro skill do?

Run a retrospective on a coding-agent session and improve the agent's environment, not the code. Combines Derby & Larsen's five-stage retrospective, Shigeo Shingo's poka-yoke (turn a mechanical mistake into a check that makes it impossible), and Matt Pocock's retro categories — navigation, automated checks, coding standards, steering files, tool economy, information access. Covers mining the session transcript for corrections, wrong turns and slow searches; deciding where each fix belongs (analyzer, hook, CI, reviewer standards, CLAUDE.md pointer, skill, doc or memory); the no-op test for stee

How do I install the Agent Retro skill?

Claude Code plugin: /plugin marketplace add vnovakovits/claude-skills. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Agent Retro popular?

It lives in vnovakovits/claude-skills, which has 5 GitHub stars and 2 forks.

Related skills

Accidental Cio

5

vnovakovits/claude-skills

Apply Scott Millett's "The Accidental CIO — A Lean and Agile Playbook for IT Leaders" (Wiley, 2024) when designing or reviewing how a technology organisation decides, funds, governs and delivers work. Covers the five components of an adaptive IT operating model (organised, work, govern, talent, lead) and its four themes (outcomes over output, intrinsic motivation, being not doing agile, manage flow not people); matching method and governance to context via Cynefin and Wardley evolution; governance as demand management + funding + measurement + decision rights, shifting from project funding to

Frontend & Design

API Design

5

vnovakovits/claude-skills

Design APIs that are clear, evolvable, and idiomatic to their protocol — covering REST (Fielding, Richardson Maturity Model, Massé), GraphQL, gRPC/protobuf, versioning strategies, pagination, error responses (RFC 7807 Problem Details), idempotency, authentication, documentation (OpenAPI), and the pragmatic trade-offs between RESTful purity and developer ergonomics. Use when designing a new endpoint, choosing between REST/GraphQL/gRPC, defining versioning, structuring error responses, deciding pagination strategy, or evaluating an existing API's evolvability.

Software Development

Architecture Decision Records

5

vnovakovits/claude-skills

Apply Michael Nygard's Architecture Decision Records (ADRs) — lightweight, immutable Markdown documents that capture significant architectural decisions, their context, and their consequences. Covers Nygard's classic format, MADR, when a decision is "architecturally significant", supersession workflow, where to store ADRs, and how to use them as living institutional memory. Use when making a significant technical decision, when joining a project and wondering "why was it built this way", when a past decision starts to bite, or when a team's tribal knowledge is bleeding away through personnel t

Software Development

AI Agents

8

Freddy-CHEN-ux/00_claude_skills

Production-grade AI agent patterns with MCP integration, agentic RAG, handoff orchestration, multi-layer guardrails, and observability (modern best practices)

AI, LLMs & Agents

Agent Centric

5

bolasblack/claude-skills

Track decisions (AGD) with validation and indexing. Use when making design decisions, recording important choices, discussing trade-offs, or when user mentions AGD or decision record.

AI, LLMs & Agents

AI LLM

8

Freddy-CHEN-ux/00_claude_skills

Production LLM engineering skill. Covers strategy selection (prompting vs RAG vs fine-tuning), dataset design, PEFT/LoRA, evaluation workflows, deployment handoff to inference serving, and lifecycle operations with cost/safety controls.

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.