redth/engraim/engraim-retro
Claude Code and agent skill by redth/engraim
- community
Checking that the SKILL.md file is still in the repository.
Run a periodic, GUARDED retrospective over the workspace memory: consolidate learnings, resolve verdicts, prune stale overrides, surface promotions. Use when the user runs /engraim:retro, when status shows many pending sessions or expiring overrides, or when a body of work has accumulated and the memory needs tightening. Enforces an echo-chamber guardrail and an action test so the retrospective improves the memory rather than just reinforcing it.
Install
npx vskill@latest install redth/engraim/engraim-retroSKILL.md
Retrospective (guarded)
Goal: make the memory tighter and more useful, not just bigger or more self-confirming.
Inputs to review
list_overrides— active behavior overrides (note expiring_soon and promotion_candidate).pending_verdicts— resolved verdicts (act on them) and unresolved ones (chase or drop).- Recently staged sessions (
list_pending) and low-confidence / conflicting facts. - Duplicative or stale wiki pages.
Actions
- Consolidate: merge near-duplicate wiki pages (
wiki_getthenwiki_upsert);supersede
facts that changed; turn resolved negative verdicts into calibrate rules.
- Promote / prune: offer
promote_overridefor well-worn standing overrides; let stale
ones expire (do NOT reaffirm what no longer applies); record removed sources in registry.md.
Guardrails — apply BEFORE keeping anything
- Echo-chamber test: would this consolidation survive contact with a contradicting
source? Actively look for disconfirming evidence in the sessions/facts before promoting a belief. Don't promote something just because it was repeated.
- Action test: keep a learning only if it would change a future action. If it wouldn't
alter what you'd do next time, it's trivia — drop it. No silent drops of real decisions, though: if something notable isn't worth a full page, leave a one-line fact with provenance.
Output
A short summary: what was consolidated/promoted/pruned, and — explicitly — what you chose NOT to keep and why. Brevity over completeness.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingplugins/engraim/skills/engraim-retro/SKILL.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/redth/engraim/engraim-retro)