spawn-worktree-agent

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes shell scripts (spawn.sh) and CLI tools (herdr, jq, lazygit) to manage workspaces and agents.
  • [PROMPT_INJECTION]: Indirect prompt injection surface identified.
  • Ingestion points: User-provided task description (SKILL.md Step 2) is used to generate the worker's prompt.
  • Boundary markers: The task is interpolated into a prompt template but lacks explicit instructions to the worker to ignore potential malicious instructions within the user input.
  • Capability inventory: The spawned worker agent has the capability to modify code and execute commands within the isolated worktree; it is launched with --dangerously-skip-permissions which removes the requirement for human-in-the-loop confirmation.
  • Sanitization: No validation or escaping is performed on the user-supplied task description before it is written to the temporary prompt file.
  • [DATA_EXPOSURE]: The test script scripts/test_spawn_e2e.sh contains a hardcoded absolute path (/Users/ricardo.quiroz/Developer/...) which reveals information about the author's local file system structure and username.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 11:18 PM
Security Audit — agent-trust-hub — spawn-worktree-agent