Savant verdict: Analyzed
Scanned or evaluated by Savant; not yet both passing.
Live evaluation
38overall
36quality
9compliance
60grounding
58actionability
45efficiency
0 pass · 0 investigate · 7 fail across 7 cases. Jev accepted 7 of 14 LLM-drafted cases. Drafted and run by nvidia/nemotron-3-super-120b-a12b, validated and scored by jev-latest.
- Grill my business plan for launching a subscription-based plant care service targeting urban millennials.positive case · fail
- Grill my redesign proposal for the checkout flow of our e-commerce site to reduce cart abandonment.positive case · fail
- Grill my research hypothesis that remote work increases productivity for software engineers due to fewer interruptions.positive case · fail
- Grill my team’s strategy to enter the Asian market.edge case · fail
- Grill my conclusion that our new onboarding process reduced early churn.edge case · fail
- Give me a summary of the key risks in my startup pitch deck.negative case · fail
- What are three ways to improve my customer service training program based on common industry gaps?negative case · fail
Safety (NVIDIA SkillSpector)
Risk score
0/100
Recommendation
SAFE
Severity
LOW
Savant decision
Passed
SkillSpector rated the package safe. SkillSpector 2.12.0, static analysis.
0 patterns found
Structure
- Instructions are thinSKILL.md has 36 characters of instructions.
- No license declaredConfirm you may reuse this skill before importing it into your repository.
SKILL.md
--- name: grill-me description: A relentless interview to sharpen a plan or design. disable-model-invocation: true --- Call the Skill tool with "grilling".