The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to execute local Python scripts that are not provided within the skill package. These scripts process user-supplied strings and external files.
Evidence: In SKILL.md, the instruction python scripts/generate_schematic.py "your diagram description" -o figures/output.png executes a script using natural language input from the user.
Evidence: In SKILL.md, the mandatory workflow for presentation review involves executing python skills/scientific-slides/scripts/pdf_to_images.py presentation.pdf review/slide --dpi 150, which processes external PDF files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and evaluate external, untrusted content such as scientific manuscripts and presentation PDFs, which could contain malicious instructions.
Ingestion points: Scientific manuscripts (manuscript and grant review) and presentation PDFs (slide decks) in SKILL.md.
Boundary markers: None defined; there are no instructions to the agent to ignore potentially embedded directives within the manuscripts being reviewed.
Capability inventory: Access to Bash, Write, Edit, and Read tools, along with the ability to execute local Python scripts.
Sanitization: No sanitization or validation of the ingested document content or the resulting script arguments (e.g., PDF filenames) is specified.

peer-review