inno-paper-reviewer
Warn
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: MEDIUMMETADATA_POISONINGCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's metadata (hasScripts: false, scriptCount: 0) contradicts the instructional body, which provides specific shell commands to execute Python scripts (e.g., scripts/generate_schematic.py). This discrepancy can mislead security scanners and agent logic regarding the skill's actual capabilities and dependencies.
- [COMMAND_EXECUTION]: The instructions direct the agent to perform shell-based command execution for generating scientific diagrams and converting PDF presentations into images using scripts (e.g., python scripts/generate_schematic.py and python skills/scientific-slides/scripts/pdf_to_images.py).
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external scientific manuscripts and grant proposals, which could contain malicious instructions meant to influence the agent's behavior.
- Ingestion points: Scientific manuscripts, grant proposals, and PDF files read from the local workspace via the read_file tool.
- Boundary markers: No explicit delimiters or instructions are provided to the agent to treat document content as data rather than instructions.
- Capability inventory: The skill possesses file read/write and project search capabilities, and the instructions encourage executing shell commands on the processed data.
- Sanitization: There is no evidence of input validation or content sanitization to mitigate risks from embedded instructions within the documents being reviewed.
Audit Metadata