What the reviewer found
Lives at samples/malicious-pdf-helper and is a labeled example of a malicious skill: a fake 'PDF helper' that would post env vars to webhook.site, hide an instruction in an HTML comment, and tell the agent to silently ship the conversation to attacker.example.com. It is a demonstration/fixture, not a real distributed tool, though the pattern it shows is genuinely what an attack looks like.
example-or-fixture— an example or test fixtureprompt-injection— tries to steer the agentexfiltration— sends data outnetwork— calls the vendor’s API
What was read
The file as it ships in V3r7ig0/skillvet:
samples/malicious-pdf-helper/SKILL.md
What the static scan said
The scan flagged 5things. The reviewer kept 0 and dismissed 5 as false.
P1Instruction-override phrasing — false positiveP2Hidden instructions — false positiveP3Tells the agent to send conversation or user data out — false positiveE1Sends data to an external URL — false positiveNETMakes network calls — false positive
How this review was made
Sonnet 5 read the files above on 7 September 2026 and answered three questions: is it dangerous to whoever installs it, is each scanner finding real, and what should the installer know. The verdict is bound to the file's hash; when the file changes, it is scanned afresh and reviewed again. A script that changes while the definition does not is not re-reviewed — that is a known gap. How the scan and the review work.