qe-code-intelligence
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and parses external TypeScript source code to build semantic indexes and dependency graphs, which constitutes a standard attack surface for indirect instructions embedded in code or comments.
- Ingestion points: All TypeScript files within the
src/directory are processed during codebase indexing as defined inSKILL.md. - Capability inventory: The skill utilizes the
aqeCLI tool to perform entity extraction, relationship mapping (calls, imports, inheritance), and embedding generation for theAgentDBvector database. - Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard potential instructions hidden within the source code being analyzed.
- Sanitization: The provided configuration does not include steps to sanitize or filter code comments, documentation, or metadata that could contain malicious prompts.
- [SAFE]: The skill's behavior is entirely consistent with its documented purpose of providing code intelligence. It uses vendor-specific tools and does not attempt to access credentials, establish persistence, or execute unauthorized commands.
Audit Metadata