The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted text provided by users for the purpose of linguistic analysis.
Ingestion points: The skill body in SKILL.md describes taking user-provided text as input for evaluation.
Boundary markers: Absent.
Capability inventory: None. The skill instructions specify it is a "find-only skill" that never rewrites content or interacts with the environment.
Sanitization: Absent.
Analysis: While the skill processes external data, it lack any exploitable capabilities such as file system access, network communication, or code execution, rendering the surface safe.
[EXTERNAL_DOWNLOADS]: The skill documentation references several external URLs for methodology attribution and academic grounding.
References include public GitHub repositories and academic resources like Wikipedia and the ACL Anthology.
These are static, informational links and do not facilitate automated software downloads or remote code execution.

check