multi-role-quality-synthesis
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process multiple 'role reports' which are external, potentially untrusted data sources.
- Ingestion points: The skill accepts role reports as input, as defined in
SKILL.mdandprompts/multi-role-quality-synthesis.md. - Boundary markers: While the prompt instructions include validation and classification rules to separate facts from constraints, there are no explicit technical delimiters (such as XML tags or unique markers) defined for the input reports to prevent the model from confusing data with instructions.
- Capability inventory: The skill does not utilize any high-risk tools such as network operations, file system writes, or arbitrary command execution.
- Sanitization: The instructions include a 'No-new-facts rule' and a 'Mandatory pre-output self-check' to ensure the output is derived strictly from the input sources, which provides a logical defense against data-driven instruction overrides.
Audit Metadata