Optimize your skills and tiles: review SKILL.md quality, generate eval scenarios, run evals, compare across models, diagnose gaps, and re-run until scores improve.
88
94%
Does it follow best practices?
Impact
88%
1.07xAverage score across 24 eval scenarios
Passed
No known issues
Optimize your skills and tiles: review SKILL.md quality, generate eval scenarios, run evals, compare across models, diagnose gaps, and re-run until scores improve.
tessl install tessl-labs/skill-optimizerThis tile combines two complementary approaches to improving your skills:
tessl eval run) — measure whether a tile makes agents better at real tasks. Scenarios are generated from the tile, an agent solves each task with and without the tile, and a judge scores outputs against a per-scenario rubric. The delta between baseline and with-tile scores shows the value-add.tessl skill review) — assess the quality of a SKILL.md file itself, without executing any tasks. A judge scores the document on fixed dimensions (completeness, actionability, conciseness, robustness) to check whether it's well-structured for routing and agent comprehension.The optimize-skill-performance-and-instructions skill combines both into a single end-to-end cycle: review → eval → improve.
| Skill | Description |
|---|---|
optimize-skill-performance-and-instructions | Runs the full optimization cycle |
setup-skill-performance | Sets up performance testing for a skill |
optimize-skill-performance | Improves how well a skill executes its tasks |
optimize-skill-instructions | Improves how reliably a skill gets invoked |
compare-skill-model-performance | Compares how a skill performs across different Claude models |