consistency-auditor
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes various external file types including LaTeX, Markdown, Python, MATLAB, CSV, and JSON files as part of its audit workflow. This creates an ingestion surface where maliciously crafted project files could attempt to influence the agent's consistency report.
- Ingestion points: The workflow scans files across the project workspace including the paper, results, methods, planning, and code directories.
- Boundary markers: No explicit markers or delimiters are specified to isolate external content from instructions.
- Capability inventory: The skill's actions are limited to reading project files and writing a local Markdown report; it lacks network access or the ability to execute code.
- Sanitization: The skill does not mention specific sanitization or filtering for the extracted text before reporting.
- [DATA_EXPOSURE]: The skill reads source code files and result artifacts to extract numerical claims and parameter values. While this involves reading the project's implementation details, it is restricted to the local workspace and does not target system credentials, private keys, or environment configuration files.
Audit Metadata