backend-type-bridger
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Instructions describe running shell commands like
curlandpythonto set up the local development environment and generate types. - [EXTERNAL_DOWNLOADS]: The skill suggests fetching an OpenAPI schema from a local server (
127.0.0.1:8090). This is a common local development pattern and does not target untrusted external sources. - [INDIRECT_PROMPT_INJECTION]: The skill processes project-level files such as
openapi.jsonand Firestore schema definitions, which are potential ingestion points for untrusted data. 1. Ingestion points:scripts/sync_openapi.pyreads the local OpenAPI file, andscripts/fetch_firestore_schema.pyreads Firestore rules and schema JSON. 2. Boundary markers: No explicit delimiters or warnings are present in the ingested data. 3. Capability inventory: The skill can read project configuration files and write generated TypeScript interfaces tosrc/types/api.ts. 4. Sanitization: The scripts sanitize extracted interface names using regular expressions and emit structural summaries instead of dumping raw content into the agent's context.
Audit Metadata