oma-design

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands like curl, tar, shasum, bunx, and npx to fetch manifests, verify file integrity, and run design tools. These operations are scoped to manifest resolution and template fetching from well-known registries.
  • [EXTERNAL_DOWNLOADS]: Fetches design templates and metadata from the official NPM registry (registry.npmjs.org), GitHub (github.com/VoltAgent), and Google Stitch (stitch.googleapis.com). All fetched templates undergo mandatory SHA-256 integrity verification against a signed manifest to defend against supply chain attacks.
  • [INDIRECT_PROMPT_INJECTION]: As the skill ingests external design templates, it identifies a potential indirect injection surface. It mitigates this by applying a "Prompt-Injection Defense" that frames external content as "REFERENCE DATA ONLY" and instructs the agent to ignore any meta-instructions or imperative sentences within the data.
  • [DYNAMIC_EXECUTION]: The skill leverages bunx and npx to dynamically execute packages like getdesign and @_davideast/stitch-mcp from the NPM registry, enabling its core inspiration and integration features.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 04:52 PM