asset-pack-intake
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from asset packs to generate intake reports and engine build plans.
- Ingestion points: The skill reads
pack.json, Tiled map files (.tmj), and feel data (.json) from an external directory provided by the user. - Boundary markers: The procedure requires running a validation script (
validate_pack.py) to check file hashes and manifest consistency before proceeding. - Capability inventory: The skill performs file system reads and uses the extracted metadata to select and sequence other skills for the project plan.
- Sanitization: The validation script explicitly identifies and blocks files with executable extensions (such as
.js,.py, and.sh) to prevent direct code execution, although it does not sanitize natural language strings within JSON files for potential prompt injection.
Audit Metadata