The Agent Skills Directory

[SAFE]: The skill's primary focus is on frontend design methodology and implementation standards. All provided instructions and scripts align with this stated purpose.
[DYNAMIC_EXECUTION]: The skill utilizes JavaScript snippets (in SKILL.md and deterministic-design/design-spatial/SKILL.md) and an external script file (deterministic-design/design-spatial/scripts/layout-audit.js) intended for execution within a browser environment via MCP tools like Playwright. These scripts perform DOM analysis (measuring layout, contrast, and accessibility) and render visual overlays for the agent to review. This is a core, documented feature for design verification and does not exhibit malicious behavior such as credential harvesting or network exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from rendered web pages and interaction traces. It proactively mitigates the risk of obedience to malicious embedded instructions by advocating for a 'separate judge' architecture—where one agent instance renders the UI and another, fresh instance audits it. This structural separation helps prevent the agent from being influenced by the content it is analyzing.

ckw-design