The Agent Skills Directory

[METADATA_POISONING]: The skill exhibits a naming discrepancy between the author ('twitchedd') and the library/repository references ('twitched', 'com.twitched.cascade'). This inconsistency is a common indicator of typosquatting or brand impersonation, which may mislead users regarding the origin and trustworthiness of the software.
[EXTERNAL_DOWNLOADS]: The skill references external packages (com.twitched.cascade and com.twitched.cascade.lite) and a GitHub repository (github.com/twitched/cascade) that do not match the author's identified resource patterns. These are treated as unverified third-party dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and prioritize instructions from external project files, specifically Packages/manifest.json and AI-GUIDE.md. This creates a vulnerability surface where malicious content within those files could override the skill's safety or logic.
- Ingestion points: Packages/manifest.json and Packages/<package>/Documentation~/AI-GUIDE.md (specified in SKILL.md).
- Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded commands within these external files.
- Capability inventory: The skill generates C# code and JSON sequences that interface with the Unity UI Toolkit runtime.
- Sanitization: The skill documents the use of SequencePreset.Validate and SequencePreset.Lint to verify the structure and behavior of generated JSON sequences.
[DYNAMIC_EXECUTION]: The skill facilitates the generation of declarative JSON animation sequences that are interpreted and executed at runtime by the Cascade library. This mechanism for turning structured data into runtime motion represents a dynamic execution pattern.

cascade