econ-audit
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from multiple sources, including source code files (.do, .R, .py), project documentation (README.md, CLAUDE.md), and PDF/Markdown pre-analysis plans. It lacks boundary markers or sanitization, creating an attack surface where an attacker could embed malicious instructions within code comments or documentation to manipulate the agent's audit findings.
- Ingestion points: File reading operations in Step 0 (target code and directories), Step 3 (cleaning code and pre-analysis plans), and Step 9 (pre-analysis plan content).
- Boundary markers: Absent. The instructions do not direct the agent to wrap external content in delimiters or explicitly ignore instructions found within the audited files.
- Capability inventory: The skill utilizes file system read and write capabilities to analyze code and generate markdown reports. No network operations were detected.
- Sanitization: Absent. No logic is provided to filter, escape, or validate the content ingested from the external files before the agent processes them for the audit.
Audit Metadata