The Agent Skills Directory

[PROMPT_INJECTION]: No attempts to override agent behavior, bypass safety filters, or extract system instructions were found. The instructional language is strictly technical and related to the stated purpose of game development.
[DATA_EXFILTRATION]: No evidence of unauthorized access to sensitive file paths (~/.ssh, ~/.aws, .env) or network exfiltration patterns. External links are limited to GitHub repositories for documentation and code reference.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns found. The skill suggests downloading a .clang-format file from the official Defold repository for code styling, which is a standard development practice and does not involve script execution.
[EXTERNAL_DOWNLOADS]: Fetches reference material and configuration files from the official Defold GitHub repository and the author's own public repositories. These sources are considered well-known and legitimate within the context of the skill's purpose.
[OBFUSCATION]: No obfuscated content, encoded strings (Base64), zero-width characters, or homoglyph substitutions were detected. The content is presented in clear, readable text.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for handling Lua callbacks within C++ extensions. These use standard Defold SDK boundary-handling functions (dmScript::LuaCallbackInfo), which represent proper practice for managing data between different execution environments.
[DYNAMIC_EXECUTION]: While the skill discusses compiling C++ extensions, this is part of the standard local build process for the Defold engine. There are no patterns involving runtime execution of dynamically generated or untrusted code.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private credentials were found in the instructions or code samples.
[PRIVILEGE_ESCALATION]: No commands for acquiring elevated permissions (sudo, chmod 777) or modifying system-level configurations were identified.

defold-native-extension-editing