source-locked-verification
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed to enhance the reliability and safety of AI outputs in high-stakes legal and evidential contexts. It implements 19 rules that explicitly forbid the agent from filling factual gaps, making assumptions, or using background knowledge without verification from authoritative sources.
- [SAFE]: The instructions mandate the use of well-known, authoritative domains for legal research, such as legislation.gov.uk and BAILII (The British and Irish Legal Information Institute). These sources are recognized as trusted authoritative repositories and their use aligns with standard professional practice.
- [SAFE]: The skill implements a robust internal verification framework (Category 8 mitigation).
- Ingestion points: The agent processes user-provided materials (documents, text, images) and online search results (SKILL.md).
- Boundary markers: The skill requires a strict five-category output classification (Rule 6) and a mandatory Source Basis table (Rule 15) to separate provided data from inferred or searched data.
- Capability inventory: The skill leverages implicit web search capabilities to perform targeted legal and factual verification (Rule 2).
- Sanitization: The skill enforces verbatim quotation rules (Rule 9) and mandatory evidential anchoring (Rule 4), preventing the agent from misrepresenting or paraphrasing source material into potentially harmful instructions or false claims.
Audit Metadata