The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection attempts was found. The instructions focus on guiding the agent to provide accurate technical and legal information about the Marey model without attempting to bypass safety filters or override system instructions.
[DATA_EXPOSURE_AND_EXFILTRATION]: No hardcoded credentials, sensitive file paths, or unauthorized network operations were detected. All external URL references target the vendor's domain (moonvalley.com), well-known services (fal.ai), or established news and press organizations for documentation purposes, which are considered safe.
[OBFUSCATION]: The content is presented in clear, plain text. There are no instances of Base64 encoding, zero-width characters, homoglyphs, or other techniques designed to hide malicious content.
[REMOTE_CODE_EXECUTION]: The skill does not contain any commands for downloading or executing remote scripts. It provides information about API endpoints (fal.ai) but does not automate their execution in a way that poses a risk.
[COMMAND_EXECUTION]: No dangerous shell commands or subprocess spawning patterns were identified. The skill is entirely informational and does not interact with the host system's shell.
[INDIRECT_PROMPT_INJECTION]: The skill does not demonstrate a vulnerability to indirect prompt injection. While it provides guidance on how to structure prompts for a video model, it does not ingest or process untrusted external data that could influence the agent's behavior maliciously.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic context injection syntax (!command) was found in the documentation.

moonvalley-marey