prompting-gpt
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSNO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of informational markdown files providing instructions to an AI agent. No executable scripts, binaries, or active code are included in the skill package.
- [EXTERNAL_DOWNLOADS]: The reference files contain links to OpenAI's developer documentation and GitHub repositories. These links target well-known, trusted official sources and are consistent with the skill's purpose of providing guidance on model usage.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted input in the form of existing prompts and tasks provided for review (SKILL.md). It mitigates injection risks by instructing the agent to establish explicit outcome contracts and use XML-style boundary markers (references/gpt-5.5.md). The skill lacks direct executable capabilities that could be exploited via injection, and does not implement automated sanitization.
- [NO_CODE]: This skill package does not contain any code or scripts. It relies entirely on natural language instructions to guide the behavior of the AI agent.
Audit Metadata