map-investors
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill ingests data from local files (
crm/companies.mdand existing domain notes) to identify target companies for portfolio overlap analysis. This untrusted data is interpolated into search queries and formatted into output files without explicit sanitization, potentially allowing malicious content in the source files to influence the agent's behavior. - Ingestion points: The skill reads
crm/companies.mdand uses the--cross-refargument to read previously generated domain notes in${WORKING_FOLDER}/domain-notes/. - Boundary markers: The instructions do not define clear boundaries or 'ignore' instructions for the content of the loaded files.
- Capability inventory: The skill possesses the capabilities to perform file reads and writes (
Read,Write), network-based information gathering (WebSearch,WebFetch), and directory management (Bash(mkdir *)). - Sanitization: No evidence of sanitization, escaping, or schema validation for the ingested CRM or cross-reference data was found.
Audit Metadata