maestro

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes git commands for worktree and branch management, and invokes shell-based test runners (such as npm test, cargo test, or pytest) identified within the target repository.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It parses external planning documents (plans, PRDs, ADRs) to extract file paths that define the scope for sub-agents. A malicious document could include sensitive system paths, potentially tricking sub-agents into unauthorized file access.
  • Ingestion points: Markdown planning documents processed by scripts/extract_files.py.
  • Boundary markers: Orchestration instructions attempt to scope sub-agents to the extracted file list, but the contents of that list are untrusted.
  • Capability inventory: Spawning new agents with modification privileges and executing test commands.
  • Sanitization: The parser excludes URLs but lacks validation for absolute paths or directory traversal strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 02:42 PM
Security Audit — agent-trust-hub — maestro