vos-port
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references official assets from the vos platform to maintain visual fidelity.
- Evidence: SKILL.md and references/elements-and-context.md direct the agent to fetch fonts from https://assets.vos.so/fonts/ and use the https://vos.so/api/fonts endpoint for catalog matching.
- [COMMAND_EXECUTION]: The skill workflow depends on executing local development commands via the @vosjs/cli package.
- Evidence: The 'Procedure' section in SKILL.md outlines steps using npx vos port inventory, npx vos build, npx vos check, and vos push to manage the program lifecycle.
- [DYNAMIC_EXECUTION]: The skill generates JavaScript functions intended for runtime execution on the target platform.
- Evidence: createContent, createTimeline, and onFrame are authored as stringified functions in config.json. references/intro-port.mjs includes a validation check to ensure no template literal interpolation occurs in the generated output, reducing the risk of code injection.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external codebases during the porting process, creating a theoretical exposure to malicious content in source files.
- Ingestion points: SKILL.md defines reading external files via npx vos port inventory.
- Boundary markers: The 'Port Grammar' in SKILL.md mandates that all user-editable content must live in config.data, isolating it from logic.
- Capability inventory: Capabilities include file writing, local command execution, and network deployment (vos push).
- Sanitization: The manual translation process required of the AI prevents direct, unexamined execution of instructions found in source project data.
Audit Metadata