v1.3.03 loop iterations

Grill Me

Most quizzing is too easy — it asks for recall of facts you just studied and confirms what you already know. Grill Me does the opposite: it finds what you don't know and forces you to confront it, increasing difficulty when you're strong and probing harder when you're shaky.

What this skill does

The point isn't to confirm what you know — it's to find the gaps. The skill assumes most self-assessments are overconfident (Dunning-Kruger is real and well-evidenced) and starts one level below where you place yourself. If your fundamentals are solid, you'll burn through the openers quickly and the difficulty escalates. If they're shakier than you thought, the gap shows up in the first three questions and the skill bores in rather than moving on.

The gap-finding algorithm is the engine. Ask a question at level N. If correct, level N+1. If incorrect, ask the same concept from a different angle — retest rather than just correcting. If still incorrect, drop to level N-1 to find where solid ground starts, then rebuild from there. Strong areas accelerate. Weak areas get probed from multiple angles until the specific point of breakdown is visible. The output isn't a score; it's a map of where your understanding actually ends.

Bloom's Taxonomy calibrates difficulty: Remember, Understand, Apply, Analyse, Evaluate, Create. Most quizzing stays at levels 1-2 — pure recall, comprehension — which is why people feel competent until they hit a real-world question that requires application or judgement. The skill spends most of its time at levels 3-5, with level 6 reserved for the synthesis round where you combine multiple concepts you were tested on into a single response. That synthesis question is where many learners discover they know the pieces but can't assemble them.

When you get something wrong, the skill doesn't lecture. It uses the Socratic method — guides you toward the answer through follow-up questions, because the cognitive work of finding the gap yourself is what makes the correction stick. There's a graduated hint system: reframe, narrow, scaffold, reveal, teach. Direct teaching only happens when Socratic guidance stalls, which means you're missing a prerequisite rather than failing the test.

The debrief is the deliverable. Solid understanding, shaky-but-present, gaps identified, untested areas, and a competence map across Bloom's levels — so you know whether the next thing to study is more practice or a different kind of practice.

When this triggers

·You've read a topic and feel like you understand it — and want to test whether you actually do
·You're preparing for an interview and need adaptive questions, not a fixed quiz
·You've been told 'you'll need to know X' and want to find the gaps before someone else does
·You're about to teach something and need to know which questions you can't answer cleanly
·You suspect you understand a topic at recall level but not at application level

Example

Trigger

User: 'Grill me on the SOLID principles. Intermediate level. ~20 minutes.'

Output

Setting expectations: I'm starting one level below intermediate. If I keep probing on a topic, I'm mapping the edge of your knowledge — not signalling that you're doing badly. Q1: Explain the Single Responsibility Principle in your own words. [You answer.] Q2: Good — but you defined what it is. Why does violating it create maintenance problems? Give me the mechanism, not the rule. [You answer.] Q3: Here's a class that handles both auth and email notifications. Which principle does it violate, and what specifically breaks when the auth library changes? [10 more questions, harder in the areas you're strong, probing different angles in the areas you're weak.] ## Debrief Solid: SRP, Open/Closed (you handled the edge case cleanly) Shaky: Liskov Substitution — you know the rule but couldn't apply it when the example wasn't textbook Gap: Dependency Inversion — fundamental misunderstanding about which way the dependency arrow points Your ceiling: Apply level. You break down when asked to evaluate trade-offs between principles. That's the next thing to study.

Get this skill + 7 more

Get the full Learning & Growth pillar (8 skills) or the complete library.

Get the full stack — $299

What you get

195-line SKILL.md, ready to drop into ~/.claude/skills/
Tested through 3 Karpathy-loop iterations (versions v1.0.0 → v1.3.0)
Triggers automatically when relevant — no command to remember
Lifetime updates as the skill is refined further