The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from rendered UIs and project configuration files.
Ingestion points: SKILL.md references reading project files in .styleseed/ and capturing screenshots of live routes.
Boundary markers: The skill instructs the agent to use perceptual scoring (e.g., "squint test") rather than automated execution from UI content.
Capability inventory: The skill utilizes Read, Write, Edit, and Bash tools across its referenced scripts and instructions.
Sanitization: Visual analysis of PNG screenshots provides a perceptual buffer between raw UI data and agent decision-making.
[DYNAMIC_EXECUTION]: The skill utilizes Playwright's page.evaluate to execute the inspectSpacing function in a browser context. This dynamic execution is limited to reading computed CSS properties and geometric data from the DOM, with safety checks implemented for input selectors.
[COMMAND_EXECUTION]: The agent is instructed to use Bash to start development servers (e.g., npm run dev) for artifact rendering. This behavior is confined to the specific project being verified and is standard for web development workflows.

ss-verify