@jamonholmgren: OK this `retro` skill is been a HUGE help for me. It's found about 15 issues so far that are, as far as I can tell, abs…
Summary
Jamon Holmgren shares that a `retro` skill (from Matt Pocock) helped him find ~15 real issues in his AI agent harness that were wasting tokens, such as linter error hints that misled agents into skipping checks and incomplete build symbols.
View Cached Full Text
Cached at: 10/04/26, 11:10 AM
OK this retro skill is been a HUGE help for me.
It’s found about 15 issues so far that are, as far as I can tell, absolutely real issues with my harness that were causing extra token usage for no reason.
As examples:
-
Linters would fail and their error message would say (among other things) “skip with –no-verify” … sometimes the agent took that literally, and just skipped checks! I removed those hints.
-
My custom Godot toolchain builds sometimes didn’t have symbols, which caused a lot of confusion in agents trying to find what function was failing. Fixed so they got included every time.
-
There wasn’t a way to see what “known issues” there were with the build, so different sessions would struggle with the same issues regularly. Added a known issues section to one of the key skills.
Thanks Matt!
Matt Pocock (@mattpocockuk): Prompt of the day - useful, or annoying?
Similar Articles
@mattpocockuk: This might be the greatest issue report in history Some folks wrote a paper about my /retro skill assessing its effecti…
An empirical arXiv study examines /retro, a skill that lets coding agents review completed sessions and propose harness changes, finding that while many proposals are generated, only some reduce repeated work on later tasks — showing harness self-evolution needs mechanisms to verify and revise changes.
@mattpocockuk: Moved my /retro skill to in-progress. You can run it after any session (or on a bunch of sessions) to try to repair its…
The article announces the update of a '/retro' skill for coding agents to in-progress, aiming to improve agent navigation, coding standards, and tool economy.
@mattpocockuk: By popular demand, I'm shipping v1.3 of my skills today. Too much hype behind /retro not to ship it. Enjoy going back t…
Matt Pocock announced v1.3 of his mattpocock/skills repo, a set of composable agent skills for coding assistants like Claude Code and Codex, featuring a new /retro skill for reviewing old transcripts, plus a Claude Code plugin and skills.sh installer for easy distribution.
@mattpocockuk: Considering adding a recommendation to /retro to turn repeated tasks into custom tooling (i.e. a CLI + skill) Talking t…
Matt Pocock 考虑在 /retro 命令中加入建议,鼓励用户将重复任务转化为自定义工具(CLI + skill),让 agent 委派任务给脚本,从而减少 token 消耗和幻觉。
@mattpocockuk: Prompt of the day: /retro read my last 10 coding agent sessions and find ways to make my repo easier to navigate. Find …
Matt Pocock shares a 'prompt of the day' that uses a retrospective review of recent coding agent sessions to spot where agents struggle to find information or rely on stale docs, arguing that improving repo navigability is an underrated way to save tokens.