vector-db-audit
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a Python script (
scripts/audit_vector.py) using the Bash tool to generate reports and CSV files. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the filesystem and database indices, which creates a potential surface for indirect prompt injection if those sources are compromised.
- Ingestion points: The skill reads from a project manifest file on the filesystem and metadata/chunks from a ChromaDB collection.
- Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content within the manifest or database entries.
- Capability inventory: The skill utilizes
Bashfor script execution andWritefor file generation (vector_audit.txt,missing_vector.csv). - Sanitization: There is no evidence of sanitization or filtering for the processed data in the provided instruction files.
Audit Metadata