living-ui-importer
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill enables importing projects from external git URLs and ZIP files, which allows the agent to pull arbitrary code from remote sources into the local environment.
- [COMMAND_EXECUTION]: The instructions provide templates for executing system commands, including package managers (
npm install,pip install) and build tools (go build,cargo build). These are used to set up and start imported applications based on detected manifests. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from imported sources, such as README files, dependency manifests, and OpenAPI specs, to synthesize requirements and operation mappings. This represents an ingestion point for potential malicious instructions, although the skill mitigates this by instructing the agent to strip credentials and treat source code as read-only reference material.
- [DYNAMIC_EXECUTION]: The agent is instructed to dynamically generate configuration files (
craftbot.jsonandoperations.json) at runtime. These files contain the command strings used to launch the application and map its internal APIs to the agent's control surface.
Audit Metadata