SkillOptimizer

Product Vision

phuryn/pm-skills2.6K installsSource at 020ee82 (opens GitHub)

44% lighter

Same results: 26 of 30 checks passed, original 25 of 30

30,050 tokens, was 53,419 across 2 tasks, 3 runs each

Check it yourself

SKILL.md keeps every line and gains one block at the end

What AIR found

  • VerifiedNo high-risk findings

    AIR's security review found nothing that needs a closer look. Reviewed Oct 1, 2026.

What changed

Original on GitHub
Files in the skill and how each changed
SKILL.mdNot in the original · Not in the lighter versionModified

Your original, with one block added at the end. Claude Code runs those commands once when the skill loads, so the agent starts with the files it always went looking for and spends no turns finding them.

Task by task

  • Refine flat vision33% lighter

    Original
    23,131 tokens
    Lighter version
    15,463 tokens
  • Create vision from scratch52% lighter

    Original
    30,288 tokens
    Lighter version
    14,587 tokens
Tokens per task, the middle of 3 runs each, on one scale.

On a smaller model

On a smaller model, the results dropped.

A separate result from the token saving: this is the cost of running the same skill on a smaller model, and the two are never added together.

Each task on the smaller model against the default model
TaskResultsCost
Refine flat visionSame results46% cheaper
Create vision from scratchLower results61% cheaper

Measured in the prepared-facts study, Sep 26, 3 runs per task, graded blind. How we measured

Embed badge
SkillOptimizer result
Markdown
[![SkillOptimizer result](/badge/product-vision.svg)](/skills/product-vision)

HTML
<a href="/skills/product-vision"><img src="/badge/product-vision.svg" alt="SkillOptimizer result" /></a>