developer-portal-design
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill outlines a design for an AI answer assistant that processes natural language input from external users. This creates a potential surface for indirect prompt injection. However, the skill incorporates mitigation strategies as part of the design requirements, including grounding answers in documentation with citations and implementing a strict fallback for unknown queries.
- Ingestion points: Natural language input through an "Ask AI" widget (SKILL.md, Step 5).
- Boundary markers: The skill mandates citations for every answer and a specific "I don't know" response for queries outside the documentation scope.
- Capability inventory: The skill itself does not include execution capabilities; it is a design framework.
- Sanitization: Implementation-level sanitization is not detailed, but the design requires RAG-based grounding to restrict the AI's response space.
- [SAFE]: The skill references several sibling skills within the
samber/namespace, which corresponds to the skill's author (Samuel Berthe). These are standard vendor resource references for modular skill architectures. - [SAFE]: External references to industry practitioners (Pronovix, Nordic APIs, Postman, etc.) and platforms (Twilio, Stripe, Mintlify, etc.) are used for architectural benchmarking and case study citations relevant to the skill's primary purpose.
Audit Metadata