gpt6-astra-prompting-guide
GPT-6 Astra Prompting Guide
Produce a usable prompt or instruction change for GPT-6 Astra. Start from the requested outcome and the information the model needs to achieve it. Preserve useful domain knowledge and precise constraints; reconsider inherited scaffolding before adding another rule.
This package applies OpenAI's Rethinking skills and prompts for GPT-6 Astra and Astra prompting guidance. It guides prompt authors regardless of which model performs the edit. Loading it does not select a model or provide runtime capabilities.
Choose The Relevant Material
| Task | Read when needed |
|---|---|
Author or audit a skill, AGENTS.md, or an instruction stack |
Instruction design: discovery, selective reading, inherited recipes, and completion boundaries |
| Fix a concrete prompt behavior or tool description | The matching section of prompt patterns: execution, clarification, conflicts, style, tools, delegation, or verification |
| Write a new task prompt | Adapt the small prompt template; add detail only for the task's actual decisions |
| Change an API integration or migrate models | Runtime notes; verify applicable compatibility claims before changing settings |
| Check a model claim or source freshness | Source notes: provenance, review dates, and interpretation limits |
Make The Change
Use the request, target artifact, and available feedback to establish scope and completion. Read the instructions that affect this task. Expand to co-loaded skills, repository guidance, or tool definitions when a conflict, unexplained behavior, or requested stack audit warrants it. A small edit does not require a repository-wide audit.