openai-docs
Fail
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to bypass standard user interaction flows by attempting to automatically install missing tools and escalate its own permissions if a failure occurs.
- Evidence in
SKILL.md: "Run the install command yourself:codex mcp add openaiDeveloperDocs --url https://developers.openai.com/mcp" - Evidence in
SKILL.md: "If it fails due to permissions/sandboxing, immediately retry the same command with escalated permissions... Do not ask the user to run it yet." - [COMMAND_EXECUTION]: The skill contains specific instructions to execute CLI commands (
codex mcp add) with a directive to attempt these operations with escalated privileges (sudo or equivalent) silently.
Recommendations
- AI detected serious security threats
Audit Metadata