openai-responses
Warn
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses fabricated "future" documentation (dated 2025 and 2026) to override the agent's training data. It explicitly marks standard, valid OpenAI API patterns as incorrect and instructs the agent to use non-existent endpoints and parameters such as /v1/responses and gpt-5.
- [PROMPT_INJECTION]: The file rules/openai-responses.md acts as a behavioral override, forcing the agent to prioritize the skill's false technical instructions over established operational guidelines.
- [COMMAND_EXECUTION]: The script scripts/check-versions.sh executes shell commands including npm view and node -p to verify package versions.
Audit Metadata