The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/brand-mod.py invokes several CLI tools including ffmpeg for GIF creation, rsvg-convert for processing SVG logos, and agent-browser for capturing preview images. These executions are performed using list arguments and are consistent with the skill's media generation purpose.
[INDIRECT_PROMPT_INJECTION]: The skill generates a web-based preview of the branded UI. The template for this preview (preview.html.tmpl) includes a simulated terminal status line that displays the text "bypass permissions on". This is a static design element for the mockup and not a functional command for the AI agent.
Ingestion points: Reads brand configuration from mod.json and brand.yaml files.
Boundary markers: Data is interpolated into HTML templates; basic validation is performed on the slug field using a regular expression.
Capability inventory: Uses subprocess.run to call external tools and performs file system writes to generate the mod directory structure.
Sanitization: Validates the slug field to prevent path traversal and ensures color hex codes match expected patterns.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to push generated code to a repository and install it using the standard claude marketplace commands, which is the expected workflow for Claude Code plugins.

brand-mod