The Agent Skills Directory

[Command Execution]: The instructions direct the agent to run a 'preflight script' from the user-context skill when local shell access is available. This is a pattern for environment initialization, which involves executing code located within the local file system.
[Indirect Prompt Injection Surface]: The skill is designed to ingest data from diverse external sources, including Storybook instances, Figma files, and codebase paths. This creates a surface where the agent processes untrusted data which could potentially contain embedded instructions.
Ingestion points: Storybook, Figma files, codebase paths, reference images, and brand assets referenced in SKILL.md.
Boundary markers: The skill contains a 'Hard boundary' instruction to not create files or start servers until outcomes are clear, but lacks specific delimiters for external content.
Capability inventory: Shell access for preflight scripts and file system access for codebase paths.
Sanitization: No explicit sanitization or validation of the ingested external content is mentioned.
[Dynamic Execution]: The skill references 'Critical Overrides' and 'User Context' via relative file paths (e.g., ../index/SKILL.md). This modular approach to loading instructions at runtime is a design pattern used to manage complex agent behaviors across multiple files.

get-context