openai-docs
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads documentation, manuals, and model metadata from official OpenAI domains (e.g., developers.openai.com). These operations are performed by internal scripts using Node.js fetch or the system curl utility.\n- [COMMAND_EXECUTION]: Local utility scripts (resolve-latest-model-info and fetch-codex-manual.mjs) are executed using shell and Node.js runtimes. The scripts also invoke the system curl command to handle network requests in environments with proxy configurations.\n- [DYNAMIC_EXECUTION]: Logic for resolving model versions and parsing documentation is implemented in bundled Node.js scripts. These scripts perform standard file system operations, such as creating and removing temporary cache files, to ensure documentation freshness.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests external content from OpenAI documentation sites, creating a surface for indirect prompt injection.\n
- Ingestion points: Markdown manuals and model guides fetched from developers.openai.com via internal scripts.\n
- Boundary markers: Not explicitly defined for the fetched external content.\n
- Capability inventory: File system access, network operations, and local script execution.\n
- Sanitization: Content is processed using regular expressions and string parsing without specific sanitization for adversarial instruction patterns.
Audit Metadata