openai-docs-skill
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- COMMAND_EXECUTION (SAFE): The script
scripts/openai-docs-mcp.shusesbashto execute commands. User-supplied arguments (queries, URLs, limits) are safely handled by passing them throughjqto construct JSON payloads, which prevents command injection vulnerabilities. - EXTERNAL_DOWNLOADS (SAFE): The skill communicates with
https://developers.openai.com/mcp. This is the official domain for OpenAI developer documentation and is considered a trustworthy source for this skill's primary purpose. - DATA_EXFILTRATION (SAFE): Network operations are restricted to the primary documentation endpoint. There are no patterns suggesting the exfiltration of local sensitive files or credentials.
- PROMPT_INJECTION (SAFE): The instructions in
SKILL.mdare focused on functional guidance for using the MCP server. No adversarial injection patterns (e.g., 'ignore previous instructions') were detected. - INDIRECT_PROMPT_INJECTION (LOW): The skill processes documentation content fetched from a remote server. While documentation could theoretically contain malicious instructions, the risk is inherent to the task of fetching external data, and the skill does not grant the fetched data any high-privilege execution capabilities.
Audit Metadata