Skills
Navigate
skills/pome-sh/skills/pome-run-task/Gen Agent Trust Hub

pome-run-task

Pass

Audited by Gen Agent Trust Hub on Oct 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the ant CLI tool from the anthropics organization's official Homebrew tap.
  • [COMMAND_EXECUTION]: The skill orchestrates the launch of external processes using the ant CLI and the pome doctor command-line utility to run agent tasks.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes and narrates external markdown content from task reports and examinee prompts, which could contain instructions intended to influence the agent's behavior.
  • Ingestion points: Data retrieved via get_report (markdown) and examinee_task.prompt (agent input).
  • Boundary markers: Explicit delimiters for the ingested content are not defined in the instructions.
  • Capability inventory: Execution of CLI tools (ant, pome), network requests to mcp.pome.sh, and the handling of session tokens.
  • Sanitization: The skill includes strong negative constraints, explicitly forbidding the writing of bearer tokens to disk, logs, or task text, and instructing the agent to "narrate, don't dump" the reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 16, 2026, 10:45 AM