agentleFS
kidowkidow/skills/review-me/SKILL.md

review-me

Skill2 starsChanged 6 months agoMITView on GitHub

Review knowledge already saved in the current repo by quizzing one item at a time with spaced-repetition scheduling, keeping all session state in a throwaway dot-folder that is deleted when done. Use when the user wants to review, quiz, or test themselves on previously learned notes, or says "review me".

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 8 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If levels.md does not exist at the repo root, offer a level check first:

    line 10
  2. Very specific:

    If yes → run place-me. If no → continue.

    line 14
  3. Distinctive:

    If I gave a topic/file as context → review exactly that.

    line 20
  4. Distinctive:

    If I gave no context:

    line 21
  5. Very specific:

    Read levels.md and propose the weakest topics first (Blank/Glimpsed before Grounded before Fluent).

    line 22
  6. Very specific:

    If levels.md doesn't exist → ask me once which topic, folder, or file pattern to review.

    line 23
  7. Distinctive:

    Don't queue the entire knowledge base by default — a growing KB makes that unusable.

    line 25
  8. Distinctive:

    If it exists (a previous session was interrupted), resume from its queue/grading data and continue.

    line 37
  9. Very specific:

    Ensure .review/ is gitignored: if .gitignore has no entry for it, add .review/.

    line 39
  10. Distinctive:

    Never commit this folder. When the session ends, delete it.

    line 40
  11. Distinctive:

    When I answer, reveal the section as the correct answer so I can self-check.

    line 46
  12. Distinctive:

    Ask me to grade: Again / Hard / Good / Easy.

    line 47
  13. Distinctive:

    When the queue is empty (every card graded Good or Easy), there is nothing left to review:

    line 59
  14. Very specific:

    Delete the .review/ dot-folder.

    line 63
  15. Very specific:

    Never modify the knowledge notes themselves — review-me only reads them.

    line 64

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Review knowledge that is already saved in this repository. Companion to teach-me: teach-me writes notes, review-me quizzes me on them. Every bit of scheduling/grading is ephemeral.

Before starting

If levels.md does not exist at the repo root, offer a level check first:

No level memory found. Want a quick level check with place-me before reviewing? (yes / no)

If yes → run place-me. If no → continue.

Scope

Decide what to review before building the queue:

  • If I gave a topic/file as context → review exactly that.
  • If I gave no context:
    • Read levels.md and propose the weakest topics first (Blank/Glimpsed before Grounded before Fluent). Ask: "Review [weak topics] today?" Stop and wait.
    • If levels.md doesn't exist → ask me once which topic, folder, or file pattern to review. Stop and wait.

Don't queue the entire knowledge base by default — a growing KB makes that unusable.

Targets

A review item ("card") is one ## section inside a knowledge note, within the chosen scope.

  • Knowledge notes live under the notes/ folder (any depth): every .md in notes/ is a note.
  • Each ## section across the matched notes becomes one card.

Session state (ephemeral)

  1. On session start, look for a .review/ folder at the repo root.
    • If it exists (a previous session was interrupted), resume from its queue/grading data and continue.
    • Otherwise create .review/ and keep all queue + grading data there.
  2. Ensure .review/ is gitignored: if .gitignore has no entry for it, add .review/.
  3. Never commit this folder. When the session ends, delete it. No review history is persisted anywhere.

Review loop

  1. Pick the next due card from the queue.
  2. Generate one question from that section's content (content-based, not just the heading). Ask it. Stop and wait.
  3. When I answer, reveal the section as the correct answer so I can self-check.
  4. Ask me to grade: Again / Hard / Good / Easy.

Each card has two learning steps. Since there is no clock, each step maps to a queue distance:

  • Again → reset to step 0; requeue after ~2 cards.
  • Hard → repeat the current step; requeue after ~4 cards.
  • Good → advance one step; requeue after ~8 cards. If already on the last step, the card graduates (removed from the queue).
  • Easy → graduate immediately (removed from the queue).
  1. Repeat. One question at a time — never ask two at once.

Ending

When the queue is empty (every card graded Good or Easy), there is nothing left to review:

  1. Tell me the session is complete with a short summary (how many cards, how many needed repeats).
  2. Update the level memory. If levels.md exists at the repo root, revise the prose summary for the reviewed topic(s) based on how this session went — what I recalled easily, what needed repeats. Update the section in place (never duplicate) and commit just that file: review: update <topic_name> level. If levels.md does not exist, skip this step.
  3. Delete the .review/ dot-folder.
  4. Never modify the knowledge notes themselves — review-me only reads them. The only file it may write is levels.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.