agents-orchestrator

Pass

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses local shell commands (ls, cat, grep) to verify the presence of project specification files and to monitor progress by counting completed tasks in the project-tasks/ directory.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of external data files.\n
  • Ingestion points: Data is read from files within project-specs/ and project-tasks/ directories as part of the orchestration workflow.\n
  • Boundary markers: The instructions do not include specific delimiters or 'ignore' commands to isolate content from these external files, potentially allowing embedded instructions to influence agent behavior.\n
  • Capability inventory: The skill spawns multiple specialist agents and provides them with content from external files; it also retains the ability to execute shell commands for state management.\n
  • Sanitization: No validation or sanitization is performed on the content of the project specification files before it is processed by the orchestrator or its sub-agents.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 11, 2026, 09:44 AM
Security Audit — agent-trust-hub — agents-orchestrator