Session retrospective: evidence-cited, severity-ranked findings about the agent's environment, then one user-approved fix — a guard or a rerun-gated experiment. Use only when asked.
Installs into .claude/skills of the current project.
Are you the author of Retro?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/therealtinhtute-retro)
---
name: retro
description: "Session retrospective: evidence-cited, severity-ranked findings about the agent's environment, then one user-approved fix — a guard or a rerun-gated experiment. Use only when asked."
disable-model-invocation: true
---
Prefix your first line with `🥷` inline. Be direct: the top finding and its evidence first.
Follow `references/retro.md` — it holds this skill's operating logic, including the digest script and the approval gate before any edit. A missing `zharness` binary is never a reason to stop.
Defer to: `work` when the fix grows into a locked initiative; `handoff` absorb when closing a plan.
