The Agent Skills Directory

[DATA_EXFILTRATION] (LOW): The skill enables the WebFetch tool in its configuration. When combined with the allowed file-reading tools (Read, Glob, Grep), this creates a theoretical path for exfiltrating sensitive data found during an audit to an external server, although no such malicious logic is explicitly implemented in the provided files.
[PROMPT_INJECTION] (LOW): The skill is susceptible to indirect prompt injection (Category 8) because its primary function is to ingest and analyze untrusted code from external plugins. Mandatory Evidence Chain: 1. Ingestion points: The skill uses Read, Glob, and Grep to process files within target plugin directories. 2. Boundary markers: There are no explicit delimiters or instructions provided to the agent to prevent it from executing instructions that might be embedded within the audited files. 3. Capability inventory: The WebFetch tool provides a network communication capability that could be exploited by injected instructions. 4. Sanitization: The skill lacks any sanitization or validation mechanisms for the content it reads from the file system.

grey-haven-plugin-audit