shadcn/improve
Has your most capable AI model write implementation plans for a cheaper model to run
shadcn/improve is a free, open-source agent skill that audits a codebase with your most capable model and writes self-contained markdown plans — exact steps, current code excerpts, and the repo's own test/lint commands as verification gates — for a cheaper model to execute. It never implements anything itself; you pick which findings get turned into plans. It records rejected findings with reasons, like "SSRF: by-design, standard proxy convention," so they don't resurface next run, and plans include STOP conditions for when reality doesn't match what was audited.
What holds up
- +Splits the expensive reasoning step from the cheap execution step, instead of paying premium rates for both
- +Plans are plain markdown — any agent or human can pick one up, not just this skill
- +Records rejected findings with reasons, so the same false positive isn't re-flagged next run
- +Plans include the repo's own test/lint commands as gates and STOP conditions for mismatches
Mind the limits
- −Only works in agents that support the Agent Skills format
- −The skill never implements anything itself — a second agent still has to run the plan
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/shadcn-improve/"><img src="https://stillworks.dev/badge/shadcn-improve.svg" alt="Picked by StillWorks" width="230" height="54"></a>