The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code as input, creating a surface for indirect prompt injection. However, the risk is mitigated as the skill only has read-only tools and no network access. * Ingestion points: Source code files read by Read, Grep, and Glob tools. * Boundary markers: Absent. * Capability inventory: Limited to read-only file access; no write or network tools are permitted. * Sanitization: Not present for input content.
[SAFE]: The skill instructions are benign and focus on providing accurate code explanations. No code obfuscation, remote code execution, or credential exposure was detected.

explain-code