Skip to content
Back to search

sprum

MarketplaceWorks with Claude CodeGitHub

Skills for Claude Code: /retro, a retrospective on the corrections in a session

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/sprum--claude-skills/badge.svg)](/listings/sprum--claude-skills)

HTML

<a href="/listings/sprum--claude-skills"><img src="/api/listings/sprum--claude-skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

/plugin marketplace add Sprum/claude-skills
/plugin install retro@sprum

Then run `/retro` (or `/retro:retro`) after a session in which you had to correct Claude. Updates: `/plugin marketplace update sprum`.

For a whole team, add this to the project's `.claude/settings.json`; Claude Code then offers the install to everyone who opens the project:

{
  "extraKnownMarketplaces": {
    "sprum": { "source": { "source": "github", "repo": "Sprum/claude-skills" } }
  },
  "enabledPlugins": { "retro@sprum": true }
}

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 16 lines
  1. {
  2. "name": "sprum",
  3. "owner": {
  4. "name": "Leo M."
  5. },
  6. "metadata": {
  7. "description": "Skills for Claude Code."
  8. },
  9. "plugins": [
  10. {
  11. "name": "retro",
  12. "source": "./plugins/retro",
  13. "description": "A retrospective on the corrections in a session: why they happened, and whether the context should change at all — and if so, where."
  14. }
  15. ]
  16. }
README.md · README · 74 lines
  1. # claude-skills
  2. Skills for [Claude Code](https://claude.com/claude-code).
  3. ## retro
  4. You corrected Claude in a session — once or several times. `/retro` goes through those corrections
  5. and asks, for each one: **why did it happen, and does anything need to change so it doesn't happen
  6. again?**
  7. For every correction it looks up, instead of guessing, whether the information already existed —
  8. in `CLAUDE.md`, the memory, the project docs, the skills that ran, the hooks — and names the cause:
  9. | Cause | Telltale sign |
  10. |---|---|
  11. | Missing | The information is nowhere. |
  12. | Not found | It exists, but somewhere Claude had no reason to read. |
  13. | Buried | It was in context, but too long, too general or without a reason. |
  14. | Ambiguous / contradictory | The instruction allows the wrong reading, or two places disagree. |
  15. | Assumed instead of asked | The decision belonged to the user. |
  16. | Mechanically checkable | A lint rule, test or hook could have caught it. |
  17. | Model error | The rule was clear and right there — and was ignored anyway. |
  18. Then it proposes a change, or none, and writes nothing until you approve.
  19. ### Why
  20. Most mistakes in agentic work are context problems — but not all of them. And every new rule has a
  21. price: it lengthens what every session loads and dilutes the rules already there. A retro that turns
  22. every correction into a new rule produces exactly the mistakes it is meant to prevent.
  23. So `retro` prefers, in this order: enforcing something mechanically (a hook, a test, a lint rule),
  24. improving an instruction that already exists, and only then adding something new — in the narrowest
  25. place that fits. If there's already a rule from an earlier correction on the same topic, that rule
  26. didn't work, and another one next to it won't either.
  27. **"Change nothing" is a complete result.**
  28. ### Example
  29. ```
  30. ### Committed to the default branch
  31. **What happened:** Claude committed straight to main instead of creating a feature branch.
  32. **Cause:** Buried — the branch rule is the 14th bullet of a long CLAUDE.md section, without a reason.
  33. **Change?** Yes — it happened before (there is a memory about it), so the existing rule isn't working.
  34. **Where:** .claude/settings.json
  35. **Change:** a PreToolUse hook that blocks `git commit` on main; delete the memory, the hook replaces it.
  36. ```
  37. ### Install
  38. ```
  39. /plugin marketplace add Sprum/claude-skills
  40. /plugin install retro@sprum
  41. ```
  42. Then run `/retro` (or `/retro:retro`) after a session in which you had to correct Claude. Updates:
  43. `/plugin marketplace update sprum`.
  44. For a whole team, add this to the project's `.claude/settings.json`; Claude Code then offers the
  45. install to everyone who opens the project:
  46. ```json
  47. {
  48. "extraKnownMarketplaces": {
  49. "sprum": { "source": { "source": "github", "repo": "Sprum/claude-skills" } }
  50. },
  51. "enabledPlugins": { "retro@sprum": true }
  52. }
  53. ```
  54. ## License
  55. [MIT](LICENSE)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.