imsg

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Recommends installing the imsg binary via a third-party Homebrew tap (steipete/tap/imsg) which is outside of the provided trusted vendor list.\n- [COMMAND_EXECUTION]: Executes the imsg CLI tool to perform system operations including listing chats, reading history, and sending messages.\n- [DATA_EXFILTRATION]: Reads sensitive message history from the macOS iMessage database (chat.db). Accessing this data requires the user to grant Full Disk Access permissions to the terminal environment.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) where malicious instructions could be received via incoming messages.\n
  • Ingestion points: Untrusted message content is ingested into the agent context via the imsg history and imsg watch commands as defined in SKILL.md.\n
  • Boundary markers: Absent. There are no delimiters or specific instructions to the agent to ignore or isolate the content of the messages from its own command logic.\n
  • Capability inventory: The skill has the capability to send messages (imsg send) and read system-level chat data, which could be exploited to exfiltrate information or influence the agent's behavior.\n
  • Sanitization: Absent. The skill does not perform any validation, filtering, or escaping of the message content before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 05:29 AM