The Agent Skills Directory

[SAFE]: The skill consists entirely of documentation and code snippets providing developer guidance. There are no executable scripts or system-level commands that pose a risk to the execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for handling user data from URL parameters, localStorage, and form inputs. While this represents the attack surface for indirect prompt injection in a live application, the provided patterns are standard for web development and do not involve the agent performing unsafe actions based on that data.
Ingestion points: Data is read from window.location.search, localStorage, and sessionStorage across various implementation files (e.g., rules/url-state.md).
Boundary markers: Data is interpolated into UI components using standard framework-specific syntax.
Capability inventory: Capabilities are restricted to browser-level storage and standard web routing via Inertia.js. No arbitrary code execution or unauthorized network access patterns are present.
Sanitization: Standard serialization methods like JSON.stringify and JSON.parse are utilized for data persistence.

durable-ui