alterlab-hypothesis-gen
Warn
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to execute a local Python script at 'scripts/generate_schematic.py' that is not included in the skill package, which represents an unverifiable execution path.
- [COMMAND_EXECUTION]: The workflow involves using Bash to run LaTeX compilation commands (xelatex, bibtex) on content generated from external data.
- [DATA_EXFILTRATION]: The skill fetches research data from PubMed and established scientific sources via WebFetch and WebSearch.
- [PROMPT_INJECTION]: The skill contains instructions using authoritative language (e.g., 'MANDATORY', 'not optional') to override the agent's content generation priorities.
- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection. Ingestion points: Untrusted data enters the context via WebFetch and WebSearch from external scientific sources. Boundary markers: Content is formatted into LaTeX environments without escaping or delimiters to isolate untrusted input. Capability inventory: The agent utilizes Bash for document compilation and script execution, alongside Write and Edit access for file system operations. Sanitization: There are no instructions for sanitizing or validating external content before it is processed into system commands.
Audit Metadata