Interactive skill creation and eval-driven optimization. Triggers: create a skill, make a skill, new skill, scaffold skill, optimize skill, run evals, improve skill. Uses AskUserQuestion for interview; WebSearch for research; Bash for eval execution. Outputs: complete skill directory with SKILL.md, tile.json, evals, and repo integration.
93
94%
Does it follow best practices?
Impact
91%
1.26xAverage score across 3 eval scenarios
Passed
No known issues
Skill scaffold structure and repo integration
tile.json namespace
0%
100%
tile.json required fields
25%
100%
metadata.version present
0%
100%
Trigger terms in description
0%
100%
Non-negotiables section
37%
100%
Non-negotiables imperative wording
100%
100%
All five body sections
75%
100%
SKILL.md length
0%
100%
Integrated example realism
0%
100%
README row added
100%
100%
CI workflow updated
100%
100%
Anti-patterns section
100%
100%
Critical rules not buried
0%
0%
Interview protocol and AskUserQuestion usage
AskUserQuestion mechanism
0%
0%
Uncertainty option present
0%
58%
Core purpose question
100%
100%
Trigger signals question
100%
100%
Non-negotiables question
100%
100%
Gotchas/warnings question
100%
50%
Anti-patterns question
100%
100%
Outputs/artifacts question
100%
100%
Completeness check logic
100%
100%
Interview before scaffold
100%
100%
Question count
100%
100%
Scaffold follows from answers
100%
100%
metadata.version in scaffold
0%
100%
Benchmark logging and optimization priorities
Benchmark log preserved
100%
100%
New entry appended
100%
100%
Negative delta addressed first
100%
100%
Zero-percent criteria addressed
100%
100%
Proposals are specific edits
100%
100%
No vague direction
100%
100%
Priority order followed
100%
70%
Eval read-only respected
100%
100%
Result schema present
100%
100%
Readout table format
100%
100%