email-triage

Warn

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute environment checks, build tools using 'bun install' and 'bun run build', and run the triage script 'digest.ts'.
  • [EXTERNAL_DOWNLOADS]: Executing 'bun install' downloads third-party dependencies from the NPM registry during the setup phase to compile the Gmail CLI tool.
  • [REMOTE_CODE_EXECUTION]: The skill runs a local TypeScript script and a compiled binary ('gmail') which allows for arbitrary code execution in the user's environment.
  • [DATA_EXFILTRATION]: Private email data is accessed via a CLI tool and transmitted to an external Telegram bot, representing a significant flow of sensitive information.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it triages untrusted email content using an LLM, which could be exploited to manipulate the agent's logic.
  • Ingestion points: Reads external email data via the 'gmail' CLI binary located in the marketplace directory.
  • Boundary markers: No explicit delimiters or instructions are used to isolate untrusted email text from system-level instructions.
  • Capability inventory: The skill has access to the 'Bash' tool, enabling command execution and script invocation.
  • Sanitization: There is no evidence of input validation or sanitization for the processed email content before it is passed to the LLM.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 27, 2026, 02:26 AM