openhound-development
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill establishes clear security boundaries for collector development. In
references/source-collection.md, there is an explicit prohibition against the collection or emission of sensitive material such as tokens or private keys. - [SAFE]: The technical guidelines in
references/preproc-lookup.mdpromote secure coding practices by recommending parameter binding for SQL queries to mitigate injection risks. - [SAFE]: Development and validation workflows in
references/validate-extension.mduse standard, industry-recognized tools (uv, ruff, mypy) within isolated temporary environments. - [INDIRECT_PROMPT_INJECTION]: The planning workflow processes external documentation and sample data provided by the user.
- Ingestion points: API documentation and sample JSON responses in
references/plan-collector.md. - Boundary markers: Not explicitly defined in the instructions.
- Capability inventory: Local file writes for model generation and subprocess execution for testing and validation.
- Sanitization: The framework utilizes Pydantic models for structured data validation and stable identifier generation.
Audit Metadata