artcraft-cli-deliver
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill installs the
artcraft-harnessdependency usingnpxfrom the npm registry and fetches machine learning models (Whisper tiny) from Hugging Face for speech-to-text functionality. - [REMOTE_CODE_EXECUTION]: The skill's bootstrap process downloads and installs the ArtCraft runtime and various domain-specific CLI tools (FilmCraft, PhotoCraft, VectorCraft, EffectCraft) required for media processing operations.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data, such as project briefs and workflow plans, which serves as a potential ingestion point for malicious instructions embedded in data.
- Ingestion points: Processes project specifications from
brief.jsonand execution plans fromplan.jsonfiles as described inreferences/brief.mdandreferences/workflow.md. - Boundary markers: Utilizes structured JSON schemas (
craft-brief/v1,craft-skill-workflow/v1) to define expected data formats. - Capability inventory: The skill has capabilities to write to the local filesystem, execute subprocesses (via domain CLIs), and perform network operations for component installation.
- Sanitization: Implements schema validation and preflight assessment steps (e.g.,
brief.py assess) before executing plans.
Audit Metadata