agent-self-scheduling

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill identifies scenarios involving the processing of untrusted external data, such as summarizing emails in the Hermes scheduler examples. This creates an attack surface for indirect prompt injection where malicious instructions embedded in the data could influence agent behavior.
  • Ingestion points: Example commands in SKILL.md suggest processing 'new emails' or 'daily standups'.
  • Boundary markers: The instructions do not explicitly provide or mandate the use of XML tags or delimiters to isolate untrusted data.
  • Capability inventory: The skill leverages bash, cron, systemd, and various agent execution tools (claude, codex, pi, hermes) which have access to the shell and filesystem.
  • Sanitization: No explicit sanitization or filtering logic is provided for the interpolated data.
  • [COMMAND_EXECUTION]: The skill provides numerous templates and patterns for executing shell commands, including while-sleep loops and crontab entries. While these are presented as standard automation techniques, they facilitate the execution of arbitrary prompts and scripts in the host environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 07:02 PM
Security Audit — agent-trust-hub — agent-self-scheduling