The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and review user-provided animation code, creating a potential surface for indirect prompt injection. Ingestion points: User-provided animation and motion code snippets. Boundary markers: No explicit delimiters for untrusted input, though the output follows a strict table format. Capability inventory: The skill explicitly disables all tool invocation in its frontmatter (disable-model-invocation: true), preventing the agent from executing commands, writing files, or accessing the network based on injected instructions. Sanitization: No input filtering is implemented.
[REMOTE_CODE_EXECUTION]: The skill documentation mentions an installation method using npx skills add to fetch the skill from a GitHub repository. This is a standard installation pattern for the platform and targets a well-known service (GitHub), making it an acceptable reference.
[SAFE]: The skill content is purely instructional and follows security best practices by limiting its own capabilities to the minimum required for code review. No credentials, persistence mechanisms, or obfuscated patterns were detected.

review-animations