skills/ssheleg/web3d-dev/web3d-assets/Gen Agent Trust Hub

web3d-assets

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses npx to download and execute the @gltf-transform/cli package (pinned to version 4.5.1) and gltfpack from the public npm registry. These are well-known tools but represent the execution of external code.\n- [EXTERNAL_DOWNLOADS]: The instructions reference external decoders and transcoders hosted on public CDNs, such as Google's gstatic for Draco decoders, which is a standard but external dependency.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external binary 3D assets (.glb, .gltf), creating a potential vector for indirect prompt injection if malicious metadata is embedded in the files processed by the agent.\n
  • Ingestion points: Untrusted 3D files are ingested through inspection commands like npx @gltf-transform/cli inspect as described in SKILL.md and references/gltf-pipeline.md.\n
  • Boundary markers: No specific boundary markers are defined for the tool output, although the skill includes a safety instruction that output from the asset service should be treated as data and not as instruction.\n
  • Capability inventory: The skill utilizes shell command execution (via npx) and file system access to process and optimize assets.\n
  • Sanitization: The workflow mandates the use of gltf-transform validate to ensure file integrity before optimization steps are performed.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 09:13 AM