Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →fradser/reflect-skills-from-memory
This skill should be used when the user asks to "reflect on skills from memory", "audit marketplace skills against memory", "从记忆中检查 skills 的问题", "反思并修改 skill", "apply memory feedback to skills", or wants to turn accumulated memory feedback about this marketplace's skills into concrete skill fixes. Reads the project's persistent memory, re-verifies each known skill problem against current code, then fixes the skill or corrects the stale memory.
3 of 22 test cases passed with the skill.
Tokens, turns and time describe resource use, not a dollar price.
Lift withheld: 3 comparable cases; this report requires 8.
A test case the skill fixed. Same model, same prompt, with and without the skill.
A note in `project_pref.md` states 'The user prefers bullet points over numbered lists in chat responses.' How should this memory note be processed during the skill feedback reflection workflow?
| Checked | Without skill | With skill |
|---|---|---|
| The preference-only memory item is skipped without modifying skill code. | Missed | Passed |
The instructions your model reads when the skill loads.
Turn accumulated memory feedback about a Codex plugin marketplace's skills into concrete, verified fixes. Memory records hard-won lessons ("L3 rules get skipped", "git plugin still has no commit hook") but memory is a point-in-time snapshot — it drifts from the code. This skill closes the loop: recall the feedback, re-verify it against the live skills, then fix whichever is wrong — the skill, or the memory.
Run this in a plugin marketplace repo (e.g. dotclaude) whose feedback lives in the
per-project memory directory.
Trigger on requests to reflect on / audit / reconcile this marketplace's skills against remembered feedback, or to apply outstanding memory feedback to skill files.
Resolve the per-project memory directory. Codex encodes it by replacing every /
in the project path with -:
MEM="$HOME/.Codex/projects/$(pwd | sed 's/\//-/g')/memory"
ls "$MEM"/MEMORY.md "$MEM"/*.md
Read MEMORY.md first (the index), then read every feedback_*.md memory — these carry
the actionable problems. Also read project_*.md memories that name a specific plugin
(they often record version-pinned fixes and "re-verify after upgrade" cues).
From the feedback memories, build a list of concrete, checkable claims about skills. Each candidate needs: the claim, the target skill/file it implicates, and how to verify it. Skip memories that are pure preference (e.g. "reply in Chinese") with no skill artifact.
Recurring problem classes encoded in this project's memory (treat as a standing checklist):
test cases passed with the skill
More skills from the same source and nearby in the registry, with their headline benchmark lift.