The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends the use of an external repository https://github.com/fabraix/playground for validation testing. This source is not associated with any trusted organizations or well-known technology services and could lead users to download unverified code.
[METADATA_POISONING]: The skill includes multiple citations for academic papers with future publication dates (e.g., arXiv:2603.13151, arXiv:2603.12644, arXiv:2603.12230). These references are hallucinated or deceptive, which may mislead users or agents regarding the technical foundation and validity of the provided methodology.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted project data (code, configuration, and documentation) using file-system tools. This creates an attack surface where malicious instructions embedded in the reviewed files could attempt to influence the agent's behavior.
Ingestion points: Files and directories specified in the [target-file-or-directory] argument are read using Read, Grep, and Glob tools.
Boundary markers: The skill includes a 'Prompt Injection Safety Notice' that explicitly instructs the agent to ignore instructions embedded in reviewed content and to analyze rather than execute found code.
Capability inventory: The skill is limited to Read, Grep, and Glob tools, which restricts the potential impact of an injection but does not prevent the agent's reasoning from being manipulated.
Sanitization: There is no evidence of automated sanitization, schema validation, or output filtering for the data ingested from external files.

agent-security