investor-call-prep
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses sensitive personal and professional data from Google Calendar, including meeting titles, event descriptions, and attendee email addresses. This information is transmitted to third-party services, including Apollo for person enrichment, Scrapegraph for website analysis, and Perplexity for firm research. While this aligns with the skill's primary function, it involves the exposure of sensitive calendar data to external research platforms.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from external sources without sufficient isolation.
- Ingestion points: Meeting titles and descriptions from Google Calendar (SKILL.md, Step 1) and content from external venture capital websites via Scrapegraph (SKILL.md, Step 3b).
- Boundary markers: Absent. The skill interpolates external data (e.g.,
{investor_email},{firm_website}) directly into command bodies and prompts without using delimiters or instructions to ignore embedded commands. - Capability inventory: Google Calendar (read), Google Sheets (write), Scrapegraph (web scraping), and Perplexity (web search).
- Sanitization: No validation or sanitization of ingested content from the calendar or scraped websites is performed before it is passed to other AI tools.
Audit Metadata