event-prospecting-pipeline

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands and local interpreters to handle data and execution.
  • Defines helper functions (state_read, state_write, state_append) that use curl to fetch/send data and python3 -c for JSON processing.
  • Uses composio-tool to check integrations, search for actions, and execute email sending via the Gmail API.
  • [EXTERNAL_DOWNLOADS]: Fetches configuration and state from a remote endpoint.
  • Interacts with a user-configured $PEPPER_CLOUD_URL to retrieve ICP definitions and existing lead pipelines.
  • [DATA_EXFILTRATION]: Transmits extracted lead data and processing logs to an external service.
  • Enriched attendee profiles, including contact information and qualification scores, are appended to a remote pipeline file hosted on the configured cloud URL.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8).
  • Ingestion points: Reads attendee lists (names, titles, and profiles) from external Luma event URLs provided at runtime.
  • Boundary markers: No explicit markers or instructions are provided to the agent to treat scraped attendee content as untrusted data during processing.
  • Capability inventory: The agent possesses the capability to write to a remote database and send emails to external recipients via Gmail.
  • Sanitization: The skill does not implement validation or escaping for the scraped attendee data before using it to generate personalized email content in Step 5.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 10:08 AM
Security Audit — agent-trust-hub — event-prospecting-pipeline