data-safety-auditor
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted source code files from user-specified directories, creating a surface for indirect prompt injection if those files contain malicious text.
- Ingestion points: File content is read in
core/DataSafetyAuditor.cjsand various detectors usingfs.readFileSyncduring the audit process. - Boundary markers: Absent; file content is analyzed without delimiters or specific instructions for the AI to ignore embedded commands.
- Capability inventory: The skill can perform recursive directory traversal, file reading, and generate reports in Console, Markdown, and HTML formats.
- Sanitization: Absent; file contents are processed raw via regular expression patterns without prior filtering or validation.
- [METADATA_POISONING]: The skill documentation includes misleading integration examples.
- Evidence:
SKILL.mdprovides a code example for adding custom rules viaauditor.rules.addRule, but theDataSafetyAuditorclass incore/DataSafetyAuditor.cjsdoes not implement this property or method.
Audit Metadata