The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external scientific manuscripts and grant proposals, which could contain malicious instructions designed to influence the agent's behavior during the review process.
Ingestion points: Processes academic research papers and grant applications provided by the user as input.
Boundary markers: The workflow lacks specific boundary markers or 'ignore embedded instructions' warnings for the processing of external document text.
Capability inventory: The agent has access to Bash (used for file conversion) and Write/Edit tools for creating reviewer reports.
Sanitization: There is no evidence of input validation or content sanitization performed on the analyzed manuscripts.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a specific Python script using the Bash tool to convert presentation PDFs into images for inspection.
Evidence: The command python skills/writing-tools/alterlab-scientific-slides/scripts/pdf_to_images.py presentation.pdf review/slide --dpi 150 is specified in SKILL.md and references/presentation_review.md as a mandatory step for reviewing slide decks.

alterlab-peer-review