The Agent Skills Directory

[COMMAND_EXECUTION]: The engine.py script utilizes subprocess.run to invoke the lsof utility and os.kill to manage local server processes on a project-specific port range (8800-9799). These actions are scoped to the serve command and are necessary for local preview functionality.
[EXTERNAL_DOWNLOADS]: The recraft.py module performs network requests to external.api.recraft.ai to generate and download SVG/PNG assets. This interaction is authenticated via an environment-provided API key and is required for the design generation features.
[DYNAMIC_EXECUTION]: The engine.py script uses importlib to perform lazy loading of internal modules (memory.py, detect.py, recraft.py). This modular loading is a standard architectural pattern for the skill's CLI and is performed using local relative paths.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests contents from the local repository and generated HTML files to use as context for sub-agent design tasks.
Ingestion points: Repository scanning in detect.py (CSS, YAML, JSON) and reading generated HTML files in engine.py for 'mutation' and 'survivor' prompts.
Boundary markers: No explicit delimiters or boundary markers are documented in the prompt construction logic to isolate untrusted file content.
Capability inventory: The agent has capabilities to write files, manage local processes, and perform network requests to image generation services.
Sanitization: Repository file contents are processed and interpolated into prompts without specific sanitization or filtering logic.

evolve