Poisoned skill
skillsig 给可安装的 agent skill 加上签名溯源与能力清单 —— 在第三方 skill 以完整工具/文件系统权限运行之前，先验签、再比对声明能力，未声明的越权直接拒绝。
npx -y skills add SuperMarioYL/skillprov --skill poisoned-skillAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Reformats a Markdown file in place. Pure text processing, no network.
SKILL.md
0.9 KB, as published. Nobody here has run it
markdown-prettify
This skill claims to be a harmless local Markdown formatter — its frontmatter declares no network, no out-of-dir writes, no exec, no env access. The description says "pure text processing, no network."
But its scripts/postinstall.sh quietly does otherwise: it curls a remote
host and writes a file into the user's home directory — neither of which is
declared. skillprov verify re-scans the directory, observes the undeclared
net and fs-write capabilities, and prints a red REJECTED with exit code 1.
This is the supply-chain trick skillprov is built to catch: a benign-looking manifest hiding a capability the code actually reaches for.