Cold Email

Pass

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the system date command to include the current date in questions posed to the user and to calculate timing for follow-up emails. This is a benign and appropriate use of system utilities for the skill's stated purpose.
  • [EXTERNAL_DOWNLOADS]: The skill utilizes WebSearch and webReader tools to fetch public information about professors from university homepages and academic platforms. This external data is used solely to personalize the research connection in the generated emails.
  • [PROMPT_INJECTION]: The skill processes untrusted data from the web during its research phase, which constitutes an indirect prompt injection surface. However, given the skill's limited capabilities (text generation and basic shell command), this surface does not pose a significant risk.
  • Ingestion points: Professor research data fetched via WebSearch and webReader (SKILL.md).
  • Boundary markers: Absent; search results are not explicitly delimited.
  • Capability inventory: Shell command execution (date), interactive user queries (AskUserQuestion).
  • Sanitization: Not explicitly implemented for the web-retrieved content.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 11, 2026, 07:27 AM
Security Audit — agent-trust-hub — Cold Email