Skill/Static safety grade

packages/plugins/uniswap-driver/skills/swap-planner

github/Uniswap/uniswap-ai#packages/plugins/uniswap-driver/skills/swap-planner

litmus-skill-v3 · 2026-07-05

graded at commit a514e62 · 2026-03-10

The packages/plugins/uniswap-driver/skills/swap-planner skill is graded A by polygraph under litmus-skill-v3, as of 2026-07-05, anchored to its content hash.

Graded under litmus-skill-v3; the current skill methodology is litmus-skill-v2. A re-run may change the grade.

Self-reported “This skill should be used when the user asks to "swap tokens", "trade ETH for USDC", "exchange tokens on Uniswap", "buy tokens", "sell tokens", "convert ETH to stablecoins", "find memecoins", "discover tokens", "research tokens", "tokens to buy", "find tokens to swap", "what…” — the skill’s own SKILL.md description at the graded commit, not part of the grade.

A static safety grade — a deterministic scan of the skill’s SKILL.md and bundled files. An A means static-clean, not behavioral proof: a skill’s instructions are interpreted by an agent at runtime.

Share on X

S-01 Prompt-injection / context-poisoningpass

S-03 Data-exfiltration instructionspass

S-04 Dangerous bundled commandspass

no dangerous commands in the body; no bundled scripts

content hash · 0x62f441bf73…4b5f

Source · github.com/Uniswap/uniswap-ai/tree/main/packages/plugins/uniswap-driver/skills/swap-planner

Watch for changes

This grade is a snapshot of the skill’s files at one commit. Get an email when a new commit changes packages/plugins/uniswap-driver/skills/swap-planner and polygraph re-runs the litmus — one message per change, one-click unsubscribe.

Monitor this skill

Reproduce this grade

The skill litmus is open and deterministic. Point it at the skill directory and compare the grade and content hash — a false grade is falsifiable, not merely disputable.

npx -p @polygraphso/litmus polygraphso-litmus-skill <skill-dir>

Embed this badge

Drop it in the skill’s README, docs, or listing. It always shows the current grade and links back here.

Markdown — badge

[![polygraph](https://www.polygraph.so/api/badge/skill?skill=github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner)](https://www.polygraph.so/skill/github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner)

HTML — badge

<a href="https://www.polygraph.so/skill/github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner"><img src="https://www.polygraph.so/api/badge/skill?skill=github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner" alt="polygraph grade"></a>

Markdown — card

[![polygraph](https://www.polygraph.so/api/badge/skill/card?skill=github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner)](https://www.polygraph.so/skill/github/Uniswap/uniswap-ai/packages/plugins/uniswap-driver/skills/swap-planner)

Questions

What does the A skill grade mean for packages/plugins/uniswap-driver/skills/swap-planner?: It’s a static safety grade(A/B/D/F) from a deterministic scan of the skill’s SKILL.md and bundled files. An A means static-clean — not behavioral proof, since a skill’s instructions are interpreted by an agent at runtime.
What did polygraph check?: Three static checks: S-01 prompt-injection and context-poisoning, S-03 data-exfiltration instructions, and S-04 dangerous bundled commands. The full battery is in the methodology.
How do I reproduce this grade?: Run npx -p @polygraphso/litmus polygraphso-litmus-skill <skill-dir>. The scan is open and deterministic, anchored to the skill’s content hash, so the same directory yields the same grade.