The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and process content from external reference URLs provided by the user, which presents a surface for indirect prompt injection. Malicious instructions could be embedded in the fetched pages to influence code generation or filesystem modifications. 1. Ingestion points: The skill utilizes the argument-hint field for reference URLs and the WebFetch tool to retrieve their content in Step 1. 2. Boundary markers: The instructions lack explicit directives for the agent to use delimiters or ignore potential instructions within the retrieved data. 3. Capability inventory: The skill is granted Bash, Write, and Edit permissions in the YAML frontmatter, allowing for file system and system-level operations. 4. Sanitization: No sanitization or validation protocols are specified for content obtained from external URLs.
[EXTERNAL_DOWNLOADS]: The skill references several well-known libraries and documentation sites within the creative coding community, such as GSAP (gsap.com), Poimandres (docs.pmnd.rs for React Three Fiber), and Lenis (github.com/darkroomengineering/lenis). These references are standard for the intended design tasks and target established service providers.
[DYNAMIC_EXECUTION]: The skill recommends using Next.js dynamic imports with SSR disabled for canvas and WebGL components. This is a standard best practice for client-side rendering of browser-specific APIs and does not involve executing untrusted dynamic code from external sources.

creative-coding