ai-knowledge-management-platform
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze architectural files and code from an external repository (TOPDEV99999/AI-Knowledge-Management-Platform). Because the skill has access to the
Bashtool and processes untrusted files, a malicious actor could embed instructions within those files to attempt to influence the agent's behavior. - Ingestion points: Untrusted repository content is read via
Read,Grep, andGlobtools, and verified using thescripts/preflight.pyutility. - Boundary markers: The
SKILL.mdinstructions explicitly tell the agent to "Confirm this exact repository" and use pinned hashes inreferences/source.jsonto verify source claims before acting. - Capability inventory: The agent has access to
Bashfor command execution and file system tools (Read,Grep,Glob) for data access. - Sanitization: The skill mitigates risks through strict behavioral instructions requiring separate human approval for database writes, credential access, cloud resource creation, and external uploads.
Audit Metadata