email-triage

Pass

Audited by Gen Agent Trust Hub on Mar 6, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of triaging untrusted external data from emails.\n
  • Ingestion points: The agent retrieves potentially malicious content from email bodies and subjects using the joelclaw email read and joelclaw email inbox commands as specified in SKILL.md.\n
  • Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its own system instructions and the untrusted content found within emails.\n
  • Capability inventory: The agent possesses the ability to archive or bulk-archive emails using joelclaw email archive, joelclaw email archive-ids, and joelclaw email archive-bulk.\n
  • Sanitization: There is no mention of sanitization or filtering of the email data before it is presented to the agent's context for decision-making.\n- [COMMAND_EXECUTION]: The skill relies on executing shell commands to interact with the Front email API via a custom CLI.\n
  • Evidence: Multiple command examples are provided in SKILL.md (e.g., joelclaw email inbox, joelclaw email read).\n
  • Context: These commands use joelclaw, which is a vendor-owned tool associated with the author (joelhooks), and are necessary for the primary purpose of the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 6, 2026, 11:01 AM