The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override agent behavior, bypass safety filters, or extract system prompts were detected in the instructions or documentation.
[DATA_EXFILTRATION]: No network operations or attempts to access sensitive system files (e.g., .ssh, .aws) were found. All URLs point to official Roblox documentation.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or credentials were found in the scripts or metadata.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. It uses standard Roblox Lua patterns for local physics simulation.
[COMMAND_EXECUTION]: No shell command execution or privilege escalation attempts (e.g., sudo, chmod) were detected.
[OBFUSCATION]: The code and documentation are clearly written and do not use Base64, hex encoding, or hidden characters to conceal intent.
[INDIRECT_PROMPT_INJECTION]: The skill focuses on internal physics logic and does not ingest or process untrusted external data that could lead to injection attacks.
[DYNAMIC_EXECUTION]: No use of loadstring, unsafe deserialization (e.g., pickle), or runtime code generation was found.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use the pre-execution command syntax (!command) in its SKILL.md file.

roblox-physics-and-constraints