The Agent Skills Directory

[SAFE]: The skill consists entirely of markdown-based instructions and reference materials. It contains no executable code, shell scripts, or commands that interact with the host system or network.
[PROMPT_INJECTION]: The skill's primary function is to analyze potentially untrusted external artifacts, which presents an inherent surface for indirect prompt injection. However, the risk is mitigated by the skill's lack of executable capabilities. * Ingestion points: Processes code, data, and analysis artifacts provided within the agent's context (SKILL.md). * Boundary markers: Utilizes structured report templates for output, though it lacks explicit instructions for the agent to sanitize or delimit untrusted input content. * Capability inventory: The skill possesses no capabilities for executing code, modifying the filesystem, or initiating network operations. * Sanitization: No specific validation or sanitization routines for analyzed artifacts are defined in the instructions.

adversarial-verifier