The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/build_gallery.py utilizes importlib.util to dynamically load and execute the Python code contained within the logo's configuration file (build.py). This allows the skill to load per-logo logic at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes external SVG data and user-provided strings which are interpolated into an HTML review gallery and a prompt for the Gemini AI model, presenting an indirect prompt injection surface.
Ingestion points: SVG path data processed in scripts/build_gallery.py and CLI arguments (logo name, description) in scripts/generate_showcase.py.
Boundary markers: Absent in the AI prompt construction; HTML templates use standard delimiters.
Capability inventory: Shell command execution via Bash, file system writes, and network access.
Sanitization: scripts/vectorize.py uses a strict regex-based tokenizer to filter SVG path commands, and scripts/build_gallery.py performs basic tag stripping on SVG inputs.
[COMMAND_EXECUTION]: The skill's workflow involves executing shell commands to run internal Python and Node.js scripts, as well as recommending the use of external image processing utilities like potrace and ImageMagick for initial tracing.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from well-known services, including Tailwind CSS and Google Fonts, to style the interactive review gallery. It also utilizes standard package managers (npm and pip) for its core dependencies.
[DATA_EXFILTRATION]: The scripts/generate_showcase.py script transmits logo image data to the Google Gemini API. This network operation is an intended feature of the skill for generating presentation renders and interacts with a trusted organization.

vectorizing-handdrawn-logos