The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface where the agent reads external data and has access to execution tools.
Ingestion points: External data enters the context via the arxiv_to_prompt tool used for fetching research papers in SKILL.md.
Boundary markers: The skill instructions do not provide delimiters or clear instructions to ignore embedded commands within the fetched arXiv content.
Capability inventory: The skill is permitted to use jupyter_execute (Python/Jupyter code execution) and latex_compile (LaTeX system execution), which provides a significant capability surface if manipulated by ingested data.
Sanitization: There is no mention of sanitization, validation, or filtering of the content retrieved from external sources before it is processed by the agent.

ml-experiment