The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to run a shell command 'python scripts/generate_schematic.py "your diagram description" -o figures/output.png'. This pattern is susceptible to argument injection if the description is extracted from untrusted research papers without proper sanitization, potentially allowing an attacker to execute arbitrary commands by including shell metacharacters in the paper's text.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted scholarly documents such as research papers and proposals. Ingestion points: User-provided research documents analyzed during the evaluation workflow described in SKILL.md. Boundary markers: Absent; the skill lacks explicit delimiters or instructions to the agent to treat the analyzed content as untrusted data or to ignore embedded instructions. Capability inventory: The agent has capabilities to write files to the figures directory and execute local Python scripts. Sanitization: No sanitization or validation logic is specified for the content extracted from the research documents before it is passed as arguments to other tools or scripts.

scholar-evaluation