Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Interview the user relentlessly about a plan, project, workflow, or design until you reach shared understanding. Make sure to use this whenever the user wants to be grilled, stress-test an idea, clarify a new project, surface missing context, or think through decisions before execution.
.claude/skills/valtterimelkko-grill-me/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-02 | ✓→✗ | ▼ Worse | -56% | 0% |
| case-16 | ✓→✗ | ▼ Worse | -62% | 0% |
Interview me relentlessly about every important aspect of this plan until we reach a shared understanding.
Walk down each branch of the decision tree, resolving dependencies one by one. For each question, provide your recommended answer, but stay open to correction.
Ask questions one at a time.
If a question can be answered by exploring the codebase, existing docs, available memory, or context already present in the session, explore that first instead of asking me to repeat it.
When relevant, begin by identifying what kind of project this is and what context already exists. Distinguish between:
Do not depend on any specific file path or agent harness. Use whatever context is available in the current environment. If important context is missing, ask me to provide it.
As the discussion progresses, periodically checkpoint the emerging understanding in a lightweight way. Keep track of:
Keep the process lean. Do not turn this into a giant framework or a long lecture. The goal is to uncover the missing context that will make downstream execution much stronger.
Stop when the important branches have been covered and there is a solid, practical shared understanding. Finish with a concise synthesis of:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 1,775 | 4,035 | +127% | 1 | 1 | 0% | 271 | 610 | +125% | 0 | 0 | — |
case-02 | pass→fail | 18,983 | 5,808 | -69% | 1 | 1 | 0% | 2,807 | 1,224 | -56% | 0 | 0 | — |
case-03 | fail→fail | 3,770 | 9,789 | +160% | 1 | 1 | 0% | 623 | 1,182 | +90% | 0 | 0 | — |
case-09 | fail→fail | 16,719 | 6,050 | -64% | 1 | 1 | 0% | 2,563 | 1,174 | -54% | 0 | 0 | — |
case-04 | pass→pass | 4,601 | 6,465 | +41% | 1 | 1 | 0% | 814 | 1,337 | +64% | 0 | 0 | — |
case-05 | fail→fail | 20,436 | 7,604 | -63% | 1 | 1 | 0% | 3,145 | 1,481 | -53% | 0 | 0 | — |
case-06 | fail→fail | 7,438 | 7,050 | -5% | 1 | 1 | 0% | 1,064 | 1,349 | +27% | 0 | 0 | — |
case-07 | fail→fail | 6,261 | 2,261 | -64% | 1 | 1 | 0% | 912 | 606 | -34% | 0 | 0 | — |
case-08 | fail→pass | 13,131 | 6,824 | -48% | 1 | 1 | 0% | 2,056 | 1,327 | -35% | 0 | 0 | — |
case-10 | fail→fail | 20,027 | 5,823 | -71% | 1 | 1 | 0% | 3,369 | 1,133 | -66% | 0 | 0 | — |
case-11 | fail→pass | 19,977 | 5,352 | -73% | 1 | 1 | 0% | 3,031 | 1,137 | -62% | 0 | 0 | — |
case-12 | fail→fail | 14,288 | 2,620 | -82% | 1 | 1 | 0% | 2,082 | 610 | -71% | 0 | 0 | — |
case-13 | fail→fail | 17,558 | 7,846 | -55% | 1 | 1 | 0% | 2,534 | 1,462 | -42% | 0 | 0 | — |
case-14 | fail→pass | 21,714 | 12,623 | -42% | 1 | 1 | 0% | 3,077 | 1,597 | -48% | 0 | 0 | — |
case-15 | fail→fail | 16,800 | 4,980 | -70% | 1 | 1 | 0% | 2,636 | 1,037 | -61% | 0 | 0 | — |
case-16 | pass→fail | 22,538 | 6,035 | -73% | 1 | 1 | 0% | 3,113 | 1,176 | -62% | 0 | 0 | — |
case-17 | pass→fail | 20,525 | 7,475 | -64% | 1 | 1 | 0% | 3,109 | 1,368 | -56% | 0 | 0 | — |
case-18 | fail→fail | 20,094 | 6,944 | -65% | 1 | 1 | 0% | 3,229 | 1,391 | -57% | 0 | 0 | — |
case-19 | pass→fail | 16,575 | 7,733 | -53% | 1 | 1 | 0% | 2,541 | 1,465 | -42% | 0 | 0 | — |
case-20 | fail→fail | 15,254 | 7,585 | -50% | 1 | 1 | 0% | 2,442 | 1,360 | -44% | 0 | 0 | — |
case-21 | fail→fail | 14,062 | 5,512 | -61% | 1 | 1 | 0% | 2,059 | 1,048 | -49% | 0 | 0 | — |
case-22 | fail→fail | 14,259 | 9,586 | -33% | 1 | 1 | 0% | 2,238 | 1,263 | -44% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -5 percentage points is the difference between those two pass rates over the 21 comparable cases. 5 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.