The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external URLs (via a browser tool) and local source code to perform design audits. This creates a surface where malicious instructions or adversarial data embedded in the target website or codebase could influence the agent's audit findings or subsequent code-fixing actions.
Ingestion points: The skill visits external websites using a browser tool and reads local repository files (routes, layout, shared components).
Boundary markers: There are no instructions provided to the agent to treat external content as untrusted or to ignore embedded instructions found within the audited data.
Capability inventory: The agent has capabilities to visit URLs, execute JavaScript in a browser, and modify local files (filesystem write access for the fixing phase).
Sanitization: The instructions do not include steps for sanitizing or validating content retrieved from external sources before processing it.
[DYNAMIC_EXECUTION]: The skill instructs the agent to execute a provided local script (references/inventory.js) within a browser environment (using a browser tool's evaluate function or the console) to collect CSS and design property data. This is a functional requirement for the design system inventory phase.

critly-audit