The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill explicitly allows the Bash tool. While the provided instructions use it for harmless directory listings, the permission itself allows for any shell command, posing a significant risk of privilege escalation or system compromise.
[DATA_EXFILTRATION] (HIGH): The toolset includes both local file reading (Read, Grep) and external network access (WebFetch, WebSearch). This creates a high-risk vector where an agent could be manipulated into reading sensitive files (e.g., credentials) and transmitting them to a remote server.
[PROMPT_INJECTION] (LOW): The skill is vulnerable to indirect prompt injection via the files it investigates. 1. Ingestion points: Reads all files in the target codebase and llmdoc/ directory via the Read, Grep, and Glob tools. 2. Boundary markers: No delimiters or protective instructions are used to separate untrusted file content from the agent's core instructions. 3. Capability inventory: High-impact tools including Bash and WebFetch are available and could be triggered by instructions embedded in the analyzed data. 4. Sanitization: None. File content is processed as-is without validation.

investigate