paw-ps-agent-product-builder
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's architecture involves ingesting data from shared memory files (specifically under
.pawbytes/prodig-suites/memory/) that are updated by various specialist agents and user interactions. Although the instructions do not specify explicit boundary markers or sanitization routines for this ingested content, the skill's capabilities are safely restricted to orchestrating other agents and managing project-specific artifacts within the project root. This structure represents a standard multi-agent coordination pattern rather than a vulnerability. - [SAFE]: The skill demonstrates legitimate functionality for product orchestration, including stage diagnosis, specialist routing, and progress tracking. It operates within a designated
.pawbytesproject directory and does not attempt unauthorized network communication, sensitive system file access, or the execution of obfuscated code.
Audit Metadata