foundry-tidas-authoring
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as source evidence and task briefs, which presents a surface for indirect prompt injection where instructions hidden within data could influence agent output.
- Ingestion points: The skill reads task JSON/Markdown files, authoring packages, and raw source evidence as described in SKILL.md and references/semantic-work.md.
- Boundary markers: Instructions explicitly require the agent to distinguish between facts, user statements, and AI assumptions, and to return unresolved questions in plain language rather than manufacturing evidence.
- Capability inventory: Agent capabilities are constrained to file authoring based on templates; direct database writes, publication commands, and validation bypasses are strictly forbidden.
- Sanitization: The skill enforces the use of predefined schemas and templates, relying on the downstream Foundry workflow for final validation and deterministic application.
Audit Metadata