get-qualified-leads-from-luma
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to run internal Python scripts for data extraction. Evidence:
python3 skills/luma-event-attendees/scripts/scrape_event.py. - [DATA_EXFILTRATION]: The skill transmits scraped attendee data to an external destination provided by the user. Evidence: Python code in Step 5 using
urllib.request.urlopen(req)to POST lead details to awebhook_url. There is no validation to ensure the URL belongs to Slack. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) as it processes untrusted data from external Luma profiles. Evidence:
- Ingestion points: Attendee bio and title data scraped from Luma and stored in
/tmp/luma_all_attendees.csv. - Boundary markers: Absent; untrusted profile data is passed as JSON directly to subagents for qualification.
- Capability inventory: Shell execution, file system access, and network POST requests.
- Sanitization: No sanitization or filtering of profile content is performed before interpolation into LLM prompts.
Audit Metadata