The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local utility scripts (run_preview.py, get_city_tokens.py, extract_variant.py) to manage design component extraction and the interactive selection workflow. These scripts facilitate communication between the AI agent and a local web browser bridge.
[COMMAND_EXECUTION]: Shell utilities including sed, eval, and mkdir are used to assemble the project files and inject design tokens. The eval command specifically processes output from the internal color extraction script.
[EXTERNAL_DOWNLOADS]: The skill uses npx to run the netlify and vercel deployment tools and references the GSAP library from Cloudflare's CDN. These are well-known and trusted external services.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it ingests untrusted user input (product name and headline) and uses it in shell command arguments.
Ingestion points: SKILL.md (Step 2 and Step 3 script blocks).
Boundary markers: Absent; user input is directly substituted into command strings.
Capability inventory: The skill uses eval, sed, npx, and gh to process input and perform network and system operations.
Sanitization: Absent; the input is not validated or escaped before being passed to the shell sink.
[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh) to automate the creation of public repositories and the configuration of GitHub Pages for the generated sites.

citycraft