The Agent Skills Directory

[PROMPT_INJECTION]: Deterministic detectors flagged override patterns in references/guides/tools-computer-use.md and references/guides/deep-research.md. Manual review confirms these snippets are part of educational examples describing how attackers might use prompt injection. They are contained within documentation blocks explaining security risks and are not malicious instructions to the agent.
[EXTERNAL_DOWNLOADS]: The documentation references official OpenAI resources, including domains like openai.com and developers.openai.com, and official GitHub repositories. These are legitimate references to technical documentation and software development kits.
[COMMAND_EXECUTION]: The guides contain numerous command-line examples (e.g., curl, npm install, pip install) and code snippets in various languages. These are intended for the user's reference when integrating with OpenAI services and are not executed by the skill itself.
[NO_CODE]: The skill bundle contains no executable scripts, binaries, or automated installation logic. It consists entirely of markdown reference files and metadata.

openai-docs