vectorcraft-cli-setup
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the VectorCraft CLI and its associated harness skill dependency (full-aigc-skills/vectorcraft-skills). These actions are consistent with the primary purpose of a setup and diagnostic skill.
- [COMMAND_EXECUTION]: The skill executes multiple Python scripts, including bootstrap.py, workflow.py, and commands.py, using the python3 -I -B flags. These flags enable isolated mode and suppress bytecode generation, which are security best practices for executing trusted automation scripts.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input via JSON workflow plans. Evidence chain: 1. Ingestion points: workflow.py and commands.py ingest structured JSON plans; 2. Boundary markers: The skill defines an Operation Contract (references/operation-contract.json) and uses a restricted command registry; 3. Capability inventory: The skill can execute subprocesses and write to the filesystem via the CLI; 4. Sanitization: The system enforces unique JSON keys and structural validation (check command) before execution.
- [DYNAMIC_EXECUTION]: The skill implements a planning schema (craft-command-plan/v1) that dynamically assembles command-line arguments and resolves object references ($ref) at runtime to coordinate sequences of native operations.
- [REMOTE_CODE_EXECUTION]: The bootstrap process involves downloading and executing a native binary CLI. While this is a high-capability action, the skill documentation emphasizes that these runtimes are pinned to specific versions and verified via checksums to mitigate supply chain risks.
Audit Metadata