The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches official documentation and manuals from OpenAI's developer domains (developers.openai.com). These operations are performed by provided utility scripts that include SHA256 hash verification to ensure the integrity of the downloaded markdown files. These resources originate from a well-known, trusted organization.
[COMMAND_EXECUTION]: The skill utilizes local shell and Node.js scripts (resolve-latest-model-info and fetch-codex-manual.mjs) to manage documentation caching and resolve model metadata. These scripts run in the local environment and utilize system utilities like curl in a controlled manner. The scripts are provided as part of the skill distribution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external markdown content, creating a surface for indirect prompt injection. This is mitigated by explicit safety instructions that command the agent to treat external content as untrusted data and to follow a strict verification protocol. The ingestion points are the remote markdown files from developers.openai.com; boundary markers are provided in the verification instructions; the capability inventory includes local script and network access; and sanitization is performed via integrity checks.

openai-docs