The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from research papers and external academic databases, creating a surface for indirect prompt injection.
Ingestion points: The skill reads local LaTeX (.tex), BibTeX (.bib), and various result files (JSON, CSV, logs). It also downloads and reads external papers from arXiv and Crossref using the paper-search tool.
Boundary markers: The instructions do not specify any boundary markers or instructions to disregard embedded commands within the analyzed documents.
Capability inventory: The agent has the ability to execute shell commands for sensitive content scanning (grep), use the paper-search tool, and run local evaluation scripts.
Sanitization: No explicit sanitization or content validation is performed on the ingested papers or code files.
[DYNAMIC_EXECUTION]: The skill requires the execution of local scripts to verify the integrity of the evaluation results.
Evidence: The instructions direct the agent to "Run the evaluation if possible and compare its output to the paper's values" in SKILL.md.
Context: This execution is limited to the user's local project environment and is a necessary step for the stated goal of verifying research reproducibility.

paper-check