agentleFS
aman-bhandariaman-bhandari/claude-code-agent-skills-framework/.claude/skills/spaced-review/SKILL.md

spaced-reviewing

Skill0 starsChanged 7 months agoMITView on GitHub

Tracks concepts the student has learned and schedules recall questions using FSRS-inspired spaced repetition. Triggers on /review-deck, quiz me, or at session start when concepts are due. Maintains a JSON deck in reflections/spaced-review-deck.json. If you cannot explain it from memory, you do not know it.

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 11 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If you can't explain a concept from memory, you don't know it.

    line 8
  2. Distinctive:

    After exercises (to add new concepts)

    line 14
  3. Very specific:

    0 -- Blank: No recall at all. → interval = 1.

    line 45
  4. Distinctive:

    After the student completes an exercise, identify key concepts

    line 53
  5. Distinctive:

    Ask: "What concepts from today should go in your review deck?"

    line 54
  6. Very specific:

    Read deck, find cards where next_review <= today

    line 59
  7. Distinctive:

    Ask the student to explain the concept from memory.

    line 61
  8. Distinctive:

    Ask a depth question: "What happens when this fails?" or "How does this work under the hood?"

    line 62
  9. Distinctive:

    Ask a production question: "How would this behave at 10x scale?"

    line 63
  10. Distinctive:

    Write updated deck back to JSON.

    line 65
  11. Distinctive:

    NEVER give the answer. If the student fails, say "review tonight" and move on.

    line 79
  12. Distinctive:

    If multiple cards in a category are struggling → flag the topic for revisiting.

    line 82
  13. Very specific:

    If any cards are due (next_review <= today), BLOCK progression.

    line 90
  14. Distinctive:

    If a card has been rated 0-2 three times in a row, the underlying exercise is flagged for re-study — the student goes BACK to that exercise.

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Spaced Review -- Retention Testing

If you can't explain a concept from memory, you don't know it. This skill ensures knowledge sticks.

Trigger

  • /review-deck or "quiz me"
  • At session start when cards are due for review
  • After exercises (to add new concepts)

Data File

reflections/spaced-review-deck.json

Card Schema

{
  "concept": "async/await",
  "category": "python",
  "topic": 1,
  "date_learned": "2026-04-08",
  "last_reviewed": "2026-04-11",
  "next_review": "2026-04-15",
  "interval_days": 4,
  "ease_factor": 2.5,
  "repetitions": 2,
  "quality_history": [4, 3]
}

FSRS-Inspired Scheduling

After each review, the student rates recall quality 0-5:

  • 5 -- Perfect: Instant, deep, with production context. → interval *= ease; ease += 0.1
  • 4 -- Good: Correct and understood, minor hesitation. → interval *= ease
  • 3 -- Acceptable: Correct but shallow, needed a moment. → interval *= ease; ease -= 0.05
  • 2 -- Struggled: Got there with hints. → interval = 1; ease -= 0.15
  • 1 -- Failed: Couldn't explain. → interval = 1; ease -= 0.2
  • 0 -- Blank: No recall at all. → interval = 1; ease = max(1.3, ease - 0.3)

Initial intervals: Day 1 → 1 day. Day 2 → 3 days. Then algorithm takes over. Minimum ease factor: 1.3 (prevents intervals from collapsing permanently).

Execution

Adding Concepts (after exercises)

  1. After the student completes an exercise, identify key concepts
  2. Ask: "What concepts from today should go in your review deck?"
  3. The student names them. Create cards with interval = 1 day.
  4. Also flag concepts they struggled with during the exercise.

Reviewing (at session start or on demand)

  1. Read deck, find cards where next_review <= today
  2. For each due card:
    • Ask the student to explain the concept from memory. No looking up code.
    • Ask a depth question: "What happens when this fails?" or "How does this work under the hood?"
    • Ask a production question: "How would this behave at 10x scale?"
  3. The student rates quality (0-5). Update card scheduling.
  4. Write updated deck back to JSON.

Analytics (on demand)

REVIEW DECK -- {date}
Total concepts: X
Due today: X
Mastered (5+ streak at quality 4+): X
Struggling (ease < 1.8): X
Categories: python: X, rag: X, agents: X, ...

Rules

  • NEVER give the answer. If the student fails, say "review tonight" and move on.
  • Questions test understanding, not memorization ("explain why" not "what is").
  • Connect recall to production: use the student's domain from memory/user_domain.md for examples.
  • If multiple cards in a category are struggling → flag the topic for revisiting.
  • Cross-reference with /assess results for comprehensive gap analysis.

Session-Start Retention Gate

At the start of every session (after the system integrity gate but before topic work):

  1. Read reflections/spaced-review-deck.json.
  2. If any cards are due (next_review <= today), BLOCK progression.
  3. The student must complete the due cards before starting new work.
  4. Cards rated 0-2 (failed or struggled) are re-queued for the next session AND trigger a "concept at risk" note.
  5. If a card has been rated 0-2 three times in a row, the underlying exercise is flagged for re-study — the student goes BACK to that exercise.

Why this gate exists: Without enforcement, students skip review. Without review, knowledge fades. Without fresh knowledge, later topics become impossible. The gate is painful on day one but saves weeks of confusion later.

Override: The student can say "skip review, I'll do it after the session" — but that must be recorded in PROGRESS.md so the pattern is visible. Repeated skipping is a red flag.

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.