The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts (scripts/generate_schematic.py and scripts/calculate_scores.py) to process data and generate outputs. These scripts take command-line arguments that are derived from user descriptions or generated content. While these are local skill components, the flow of user-influenced strings into shell commands presents a potential injection risk if the scripts do not properly sanitize arguments.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to analyze and evaluate external scholarly work (research papers, proposals, abstracts). These documents constitute untrusted data that could contain malicious instructions designed to subvert the agent's logic or evaluation criteria.
Ingestion points: Scholarly documents (PDF content, text, or abstracts) provided by the user are ingested into the agent context in SKILL.md Step 1.
Boundary markers: The instructions do not define explicit delimiters or 'ignore embedded instructions' warnings for the analyzed content.
Capability inventory: The skill can execute local Python scripts (scripts/generate_schematic.py, scripts/calculate_scores.py) and write files to the disk (figures/ directory, report.txt).
Sanitization: No evidence of content sanitization or escaping is present in the instructions prior to analysis or script input construction.

scholar-evaluation