agent-teams

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_CONTEXT_INJECTION]: The skill uses shell execution markers in the main documentation to display active Claude processes and git status. These commands are non-privileged and used solely for monitoring the local environment.
  • [COMMAND_EXECUTION]: The orchestration patterns include the use of the --dangerously-skip-permissions flag for sub-agents to enable autonomous operation. This is a standard feature for the managed CLI tool within this specific orchestration context.
  • [INDIRECT_PROMPT_INJECTION]: The skill handles data from parallel agents, creating a potential surface for indirect prompt injection. 1. Ingestion points: Monitoring of .agent-status files and git log output from sub-agent worktrees. 2. Boundary markers: The instructions mandate strict file isolation and individual context windows for each agent session. 3. Capability inventory: Shell command execution, git operations, and local file management delegated to the worktree-manager skill. 4. Sanitization: Relies on the Team Lead agent to perform manual verification of code changes and test results prior to merging work.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 09:56 AM
Security Audit — agent-trust-hub — agent-teams