hve-builder
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFE
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to ingest and process prompt-engineering artifacts (targets), which enters the agent context. This presents a potential surface for indirect prompt injection if the artifacts contain malicious instructions. However, the skill incorporates specific mitigations, such as explicit instructions in SKILL.md and its references to treat all external content strictly as data rather than instructions, and the use of isolated subagents to separate authoring from review.
- Dynamic Subagent Orchestration: The workflow dispatches specialized subagents like HVE Artifact Author and HVE Artifact Reviewer to perform isolated tasks. While delegating capabilities to subagents increases the complexity of the trust model, it is used here to provide context isolation and ensure that sensitive operations like file modification are subject to independent verification.
- Safety and Enforcement Controls: The skill follows industry best practices by instructing the agent to keep secrets out of artifacts and to verify third-party extensions. It also establishes a safety boundary for host-provided extensions, ensuring they cannot redirect the workflow or weaken the skill's base security rules.
Audit Metadata