vectorcraft-cli-selection
Warn
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads a binary executable (
vectorcraft-cli) from thestorytold/vectorcraftrepository on GitHub and executes it locally through a bootstrap script during initialization. - [EXTERNAL_DOWNLOADS]: Fetches executable binaries and desktop application packages (DMG/archives) from remote sources that are not part of the pre-approved trusted organization list.
- [COMMAND_EXECUTION]: Executes multiple Python scripts (
bootstrap.py,cli.py,workflow.py, etc.) and shell commands to coordinate the CLI tool's operations and manage vector project files. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Ingests external
.vectorcraftproject files andmanifest.jsonmetadata for processing and modification. - Boundary markers: Utilizes SHA-256 hash verification to ensure file integrity, although it lacks specific protection against malicious instructions embedded within vector object text or metadata fields.
- Capability inventory: The skill possesses powerful capabilities including subprocess execution, filesystem write access, and network download operations.
- Sanitization: Employs basic file integrity checks but does not describe any sanitization or filtering of content for potential prompt injection attacks contained in vector data.
- [DYNAMIC_EXECUTION]: Orchestrates operations by dynamically loading and executing scripts from a harness directory based on user-provided plans and templates.
Audit Metadata