The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection because it is designed to ingest and analyze untrusted external data (scholarly papers, research proposals, and abstracts) as described in SKILL.md (Step 1 & 2).
Ingestion points: Research papers, literature reviews, and abstracts provided by the user for evaluation.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the evaluation workflow.
Capability inventory: The skill includes file-writing capabilities through scripts/calculate_scores.py (via the --output flag) and references a script scripts/generate_schematic.py for image generation.
Sanitization: There is no evidence of sanitization or filtering applied to the ingested scholarly content before processing.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute local Python scripts (scripts/calculate_scores.py and scripts/generate_schematic.py). While calculate_scores.py is provided and found to be safe, scripts/generate_schematic.py is referenced for execution (e.g., python scripts/generate_schematic.py "your diagram description" -o figures/output.png) but is missing from the skill package files, rendering its behavior unverifiable.

scholar-evaluation