Skip to content

feat(thinking-tools): add quiz-me skill - #13

Open
mickzijdel wants to merge 1 commit into
mainfrom
claude/youthful-noether-gnue67
Open

mickzijdel wants to merge 1 commit into
mainfrom
claude/youthful-noether-gnue67

Conversation

@mickzijdel

Copy link
Copy Markdown
Owner

Summary

  • Adds a quiz-me skill to thinking-tools (v1.2.1 → v1.3.0)
  • Inspired by Geoffrey Litt's feed item (2026-06-28): have coding agents quiz developers on their modifications to build genuine comprehension
  • Active recall beats passive reading: instead of the agent explaining, the developer is questioned on key decisions, patterns, and edge cases in the change

What the skill does

  1. Surveys the current diff — identifies 3–5 key decisions, mechanisms, and non-obvious constraints introduced
  2. Constructs 4–6 targeted questions covering five types: recall (basic concept), reasoning (why this approach), implication (what follows), trace (follow the execution), and edge case (boundary condition) — at least one trace and one implication required
  3. Presents questions one at a time — no list upfront (prevents skimming ahead); waits for each answer
  4. Scores and explains each answer — confirms what was right, fills gaps, explains the principle behind wrong answers so the mistake doesn't recur
  5. Ends with a comprehension summary and any follow-up action (e.g. "worth adding a comment to X")

What changed

  • plugins/thinking-tools/skills/quiz-me/SKILL.md — new skill
  • plugins/thinking-tools/.claude-plugin/plugin.json — version bump, description updated, comprehension/learning keywords added
  • plugins/thinking-tools/README.md — new row in skills table + usage example
  • README.md — thinking-tools entry updated to include quiz-me

How it's distinct from existing skills

Skill Focus
grill-me Attacks the code for bugs, security holes, edge cases
but-for-real Verifies the agent's claims about the change
code-simplifier Style, clarity, and refactoring
self-rate Grades the agent's own work quality
quiz-me Builds the developer's comprehension of what the agent just did

Test plan

  • uv run pytest -q — no test changes needed; quiz-me is a pure skill
  • Verify skill table in plugins/thinking-tools/README.md renders correctly
  • Invoke /quiz-me after a small implementation and confirm it goes straight to Q1 without preamble

Source

Geoffrey Litt feed item 2026-06-28 (https://x.com/geoffreylitt/status/2071319131144790064): "Have coding agents quiz developers on their modifications and construct test environments to verify understanding of code alterations."

🤖 Generated with Claude Code


Generated by Claude Code

Inspired by Geoffrey Litt's feed item (2026-06-28): have coding agents quiz
developers on their modifications to build genuine comprehension via active recall.

Adds quiz-me to thinking-tools (v1.2.1 → v1.3.0): surveys the current diff,
constructs 4–6 targeted questions (recall, reasoning, implication, trace,
edge-case), presents them one at a time, scores and explains each answer,
and ends with a comprehension rating. Distinct from grill-me (attacks code for
bugs), but-for-real (verifies agent claims), and code-simplifier (style).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant