agent-bridge
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to scan an entire codebase to identify application actions, logic, and side effects. This ingestion process is susceptible to indirect prompt injection if the source code or associated documentation contains malicious instructions intended to mislead the agent during the discovery and planning phase.
- Ingestion points: Automated codebase scan described in
references/step-1-discover-actions.md. - Boundary markers: Absent; the skill does not define specific markers to separate untrusted codebase content from its internal instructions.
- Capability inventory: The agent has the capability to write new API routes, refactor existing code, and suggest database migrations across the application.
- Sanitization: The skill relies on interactive user interviews (
AskUserQuestion) and manual review of theAGENT_ACTION_PLAN.mdfile rather than programmatic sanitization. - [COMMAND_EXECUTION]: The skill instructs the agent to implement infrastructure changes, generate database migrations, and modify application logic. These actions involve significant changes to the operating environment. The skill proactively mitigates these risks by enforcing localhost-only defaults, requiring secret-based authentication, and recommending that all write actions go through a human approval dashboard.
Audit Metadata