The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection patterns. The instructions focus exclusively on providing a technical implementation for scroll-driven animations without attempting to override agent behavior or safety filters.
[DATA_EXFILTRATION]: No sensitive file paths or credentials were found. There are no network operations targeting untrusted domains. The only external references are to well-known, industry-standard libraries (GSAP and Lenis) and standard HTML media attributes.
[OBFUSCATION]: The code and documentation are provided in clear, plain text. No Base64, hex-encoded strings, or hidden Unicode characters were detected in the scripts or metadata.
[REMOTE_CODE_EXECUTION]: The skill uses established packages (gsap, lenis) but does not attempt to download or execute code from untrusted remote sources. It provides a static reference implementation rather than dynamic script loading.
[DYNAMIC_EXECUTION]: The JavaScript uses standard DOM manipulation (e.g., document.createElement) to create progress indicators. No unsafe evaluation methods like eval() or exec() are present.
[INDIRECT_PROMPT_INJECTION]: While the skill processes data-act and data-anim attributes, this is performed locally for animation purposes and does not represent a significant attack surface for external instruction ingestion. The logic is constrained to UI choreography.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use the !command syntax or any other dynamic context injection patterns. References to shell-like commands in EVAL.md are for manual/automated auditing of the code structure and are not executed during skill load.
[PRIVILEGE_ESCALATION]: No administrative commands, privilege elevation requests (e.g., sudo), or persistent access mechanisms (e.g., cron jobs, registry keys) were identified.

scroll-horizontal-act-sequence