figcraft
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a local loopback socket (
127.0.0.1:7979) to send JSON-based instructions to the FigCraft application. It includes documentation for usingfigcraft-cliand standard shell socket redirection to drive the interface. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from external files and application events which could be manipulated to influence agent behavior.
- Ingestion points: Ingests design files (SVG, PNG, PDF,
.figcraft) and monitors comment activity via theevents.pollmethod. - Boundary markers: No explicit delimiters are provided in the protocol instructions to isolate data from potential instructions within the ingested content.
- Capability inventory: The skill possesses the ability to execute design engine commands, manipulate UI state, and perform filesystem writes through screenshot and render methods.
- Sanitization: The documentation notes that the engine enforces strict parameter validation and that programmatic calls are prohibited from opening interactive dialogs.
- [EXTERNAL_DOWNLOADS]: Refers to and downloads protocol documentation from the
storytold/figcraftrepository on GitHub. This repository is cited as the official source for the application's automation specifications.
Audit Metadata