<- all tokdocs

Matt Pocock's /retro Skill: What It Does and What It Doesn't

By , founder of Mavric and Apso

View on TikTok ->

/retro is a skill in Matt Pocock's mattpocock/skills repo that reviews a coding agent session and suggests repo improvements, such as a new lint rule or a line in CODING_STANDARDS.md. In a TikTok from @mattpocockuk he says it reviews "your previous sessions" and "will add" the fixes. The skill's own files say it reads the current session by default and only proposes changes, then waits for you to pick one.

What is the /retro skill?

The on-screen page describes it as "Look back on a session and suggest repo improvements." The pitch in the video is a question: when an agent makes a mistake, "how do I make that mistake harder for the agent to make next time?" His answer is retro, which turns mistakes into "automated checks" or standards files so that "code review agents catch those mistakes before they go to production."

Install it with the command shown on screen:

npx skills@latest add mattpocock/skills --skill=retro

Then type /retro in your coding agent. That command uses the Vercel Labs skills CLI, which installs into Claude Code, Cursor, Codex, GitHub Copilot, Windsurf and others. The repo README also offers claude plugin install mattpocock-skills@claude-plugins-official.

Is /retro new in v1.3?

He calls it "fresh in my skills repo for V1.3." Release v1.3.0 was published on 2026-10-04, nine days before the video, with the changeset "Graduate retro into the Engineering bucket, so it ships in the Claude Code plugin, gets a docs page, and is routed by ask-matt as the last step of the main flow, after code-review."

The skill itself is older. It was first committed on 2026-08-24 as an in-progress skill, and PR #1120 moved it to skills/engineering/retro/ on 2026-09-29. What v1.3 changed was its status. It now runs as the last step of the chain grill-with-docs → to-spec → to-tickets → implement → code-review → retro.

Does /retro review previous sessions or the current one?

The video says retro "conducts a retrospective on your previous sessions." SKILL.md says: "If the user doesn't specify a session, default to the current one." The docs page agrees: "It reads the session's own record (the current one by default, or one you point it at in the session logs)."

It reads one session per run. The docs FAQ says "It sees one session," and gives that as the reason it "cannot tell you that a rule now fires too often or that the bug behind it is gone." The docs also call the current session "the best case, because the struggles are still in the context window." You can point it at an older session log, but that is the fallback.

Does /retro add rules automatically?

No. The video says retro "will add them into steering files like codingstandards.md." The last step of SKILL.md reads "Present these candidates to the user, in order of severity." The docs FAQ is direct: "It waits. retro only proposes. Nothing changes until you pick a candidate, so it edits nothing by hand and installs no hook automatically."

The docs say this is deliberate, and quote a user who had been "burned by auto-hooks that blocked good changes." There is no dry-run mode, so test a proposed check against the repo before letting it block merges.

The agent also cannot start it on its own. The frontmatter sets disable-model-invocation: true, which per the Claude Code skills docs means the user must type /retro.

Checks vs coding standards: where do retro's fixes go?

The video's two destinations match the skill. SKILL.md tells it to classify each violation first. A mechanical one (a banned API, an import shape, a file-location rule) "gets a deterministic check, full stop," meaning a linter rule, pre-commit hook or CI job. CODING_STANDARDS.md is reserved "for genuine judgement calls." The docs give the reason: "a check can fail and a sentence in a standards file can't."

The skill also treats a missing guardrail as a problem: "A repo with no guardrail (no pre-commit hook and no CI job running its lint/typecheck/test command) is itself a finding."

The code-review handoff in the video is accurate. The code-review skill reads CODING_STANDARDS.md when it exists. No skill ships the file. Retro proposes creating it the first time a judgement-call rule comes up.

One wording fix. The video calls CODING_STANDARDS.md a steering file. In the skill's terms, steering files are CLAUDE.md and AGENTS.md, which load into every agent's context and "should be used incredibly sparingly." The docs say retro "pushes back on" adding lines to CLAUDE.md "more than anything else."

What is a Claude Code skill?

Per Claude Code's skills docs, a skill is a SKILL.md file with YAML frontmatter and Markdown instructions, and the folder name becomes the slash command. Anthropic announced Agent Skills on October 16, 2025, and agentskills.io describes the format as an open standard used by Claude Code, Cursor, Codex and others.

What are /retro's known weaknesses?

The video offers testimony ("Retro has been absolutely invaluable for me") and shows no run of the skill. The repo docs are more candid and list three limits:

  • Generic advice. The docs call this "the strongest criticism of it," quoting a user who said the AI "invents generic advice to satisfy the retro categories." The advice to users is "Discard any candidate you can't trace to a specific moment."
  • Rule pile-up. Retro "does not audit the lint rules, hooks, or CI jobs it proposed last month." The docs call this "its weakest spot."
  • No memory. Retro "is not a memory system. It does not store what happened."

Bottom line

/retro is a real, documented skill that turns one session's mistakes into a ranked list of proposed checks and standards. It reads the current session unless you point it elsewhere, and it changes nothing until you choose a candidate. Prefer the checks it proposes over new lines in a standards file, as the skill itself does.

Sources

Published October 13, 2026. Based on a TikTok by @mattpocockuk. Full video notes (markdown)