vectorcraft-cli-project
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill initiates the download and installation of the VectorCraft CLI and associated harness components (e.g.,
vectorcraft-harness). These resources originate from the vendor's repository and the installation process is managed by a bootstrap script that performs integrity checks via aruntime.lock.jsonfile. - [COMMAND_EXECUTION]: Core functionality is implemented through the execution of Python scripts (
bootstrap.py,cli.py,workflow.py,commands.py). These scripts are run using security-hardening flags (-I -B) to isolate the execution environment and prevent interference from external Python configurations. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data including vector project files (
.vectorcraft), SVG, and PNG assets, as well as user-provided JSON operation plans. This represents a potential attack surface for indirect prompt injection. - Ingestion points: Operations read content from external project files and media assets specified in the
operationsandassetsfields of the JSON plans. - Boundary markers: The skill relies on structured JSON schemas (
craft-command-plan/v1) to delimit commands and parameters, reducing the risk of instructions being misinterpreted as data. - Capability inventory: The skill facilitates local file operations (reading/writing projects) and subprocess execution of the authorized VectorCraft CLI.
- Sanitization: Input parameters are validated against command schemas, and the system employs integrity locking for dependencies and assets.
Audit Metadata