The Agent Skills Directory

[SAFE]: The skill demonstrates a well-defined operational scope, using relative paths to access internal reference files for its analysis modes. It does not request network access, file system writes, or the execution of shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied materials including files, screenshots, and pasted text (Ingestion point: SKILL.md). It explicitly includes boundary markers by referencing safety guidelines (Boundary: 06_external-material-safety.md). The skill lacks dangerous capabilities such as subprocess calls, network operations, or file-writing tools (Capability inventory: none). The inclusion of safety protocols serves as the primary sanitization mechanism (Sanitization: safety references). The risk is assessed as safe due to these mitigations.

eva-lens