The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill can fetch content from external URLs or local files provided by the user using WebFetch to assist in drafting documentation. This creates a surface where external data could potentially contain instructions aimed at influencing the agent's output.
Ingestion points: Step 4a reads external documents via URLs or local file paths in SKILL.md.
Boundary markers: The skill contains explicit instructions to treat the fetched content as data only and to ignore any imperative commands within it.
Capability inventory: The skill can perform file writes, execute shell commands to open a browser, and invoke subsequent skills.
Sanitization: Safety is maintained through specific instructions for the agent to distinguish data from instructions.
[Automated Command Execution]: During the design review phase (Step 8.0), the skill uses the Bash tool to automatically run shell commands (such as open, xdg-open, or start) to display a generated HTML preview in the user's browser. This is an automated execution of system commands based on generated file paths, intended to streamline the visual review process.

design-native-extension-feature