orchestrate

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill interpolates user-provided input through the $ARGUMENTS variable directly into the orchestration workflow without protection.
  • Evidence: The $ARGUMENTS placeholder in SKILL.md captures raw user input which is then used to generate prompts for sub-agents via the Agent tool.
  • Impact: The lack of boundary markers (such as XML tags or triple quotes) and instructions to ignore embedded commands makes the orchestrator susceptible to following instructions hidden within the user's task description.
  • [PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection due to its core capability to delegate tasks.
  • Ingestion points: User input enters the context through $ARGUMENTS in SKILL.md.
  • Boundary markers: Absent. There are no delimiters surrounding the user input to distinguish it from the skill's instructions.
  • Capability inventory: The skill utilizes highly privileged tools including Bash, Write, Edit, and Agent (which can spawn and instruct further agents).
  • Sanitization: Absent. The skill does not perform any validation or escaping of the user-provided task description before processing it.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 12:05 AM
Security Audit — agent-trust-hub — orchestrate