skills/irpsv/ai-bro/bro-do-it/Gen Agent Trust Hub

bro-do-it

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of plans, specifications, and code context, which are passed to subagents for execution. \n- Ingestion points: Untrusted data enters the agent context via placeholders in subagents/developer-prompt.md and subagents/verifier-prompt.md. \n- Boundary markers: Uses XML-style tags to delineate external context and provide scoping instructions to the subagents. \n- Capability inventory: Allows for file modifications, git commits, and shell command execution for testing. \n- Sanitization: Relies on LLM-based filtering and explicit escalation rules in references/autonomy-and-stop.md which requires manual intervention if a security threat is identified. \n- [COMMAND_EXECUTION]: The verifier subagent is instructed to execute shell commands to run automated tests, static analysis, and builds as part of the verification process in subagents/verifier-prompt.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 11:42 AM
Security Audit — agent-trust-hub — bro-do-it