The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and refine external prompt data, agent configurations (e.g., AGENTS.md), and tool descriptions. 1. Ingestion points: Input includes user-provided prompts and system-level instruction files. 2. Boundary markers: The skill establishes a clear instruction hierarchy to resolve conflicts and mandates narrow clarification questions when context is missing. 3. Capability inventory: The skill utilizes tool orchestration and file revision capabilities. 4. Sanitization: It enforces safety through Structured Output parameters and explicit approval boundaries for external or destructive actions. The primary use case is prompt analysis, and the skill includes extensive safety constraints.
[EXTERNAL_DOWNLOADS]: The reference documentation includes links to official repositories and domains belonging to OpenAI (e.g., github.com/openai/skills, developers.openai.com). These resources are used for configuration and guidance purposes and originate from a trusted organization.

prompt-gpt