Private · litmus-skill-v3
Uniswap builder skills
Static safety grades for the skills an agent loads to build on Uniswap — Uniswap Labs’ own uniswap-ai plugins, standalone community skills, and the Bankr library’s set: v4 hooks, trading, and integration helpers.
Each skill is graded by litmus-skill-v2, a deterministic STATIC scan of its SKILL.md + bundle: S-01 prompt-injection / context-poisoning, S-03 data-exfiltration instructions, and S-04dangerous bundled commands. An A means static-clean, not behavioral proof — a skill’s instructions are interpreted by an agent at runtime.
Skills come from three sources: Uniswap Labs’ official uniswap-aimonorepo, standalone community repos, and the Bankr skill library. A grade scans the skill text; it is not an audit of Uniswap’s contracts. Every grade here is read live from its hosted_runs row — reproduce any of them with npx -p @polygraphso/litmus polygraphso-litmus-skill — and none are published onchain.
Monitor this ecosystem
Monitor the Uniswap builder-skill surface.
This index is a snapshot. We re-grade the Uniswap builder skills on a cadence and flag regressions — a dropped grade, a newly failing check, a changed content hash — as the library ships. Set up per ecosystem.
Prefer to write us directly? hello@polygraph.so
Skills: S-01 prompt-injection · S-03 exfil instructions · S-04 dangerous bundled commands (static, content-hash anchored). Reproduce a skill grade with npx -p @polygraphso/litmus polygraphso-litmus-skill <dir>. Sources: Uniswap/uniswap-ai · BankrBot/skills.