Use progressive disclosure for complex skills with multiple workflows
For skills with many workflows, use a minimal root document that links to supporting details.
For skills with multiple workflows, make the root document a minimal router that points to supporting docs and scripts.
This approach provides the model with just enough guidance to know where to find information without forcing it to process irrelevant details. It conserves context and improves efficiency by only loading necessary instructions.
A root skill document that briefly describes its purpose and then links to separate documents for 'Database Migrations', 'API Integration', and 'Testing Procedures'.
Implement progressive disclosure for complex AI skills by using a main document as a router to specific sub-documents. This prevents overwhelming the model with unnecessary information, saving context and improving efficiency.
- Rethinking skills and prompts for GPT-6 Astra ↗OpenAI Developers Blog · OpenAI Developers Blog · Sep 12, 2026 · paraphrased
First seen on Subvolts Sep 12, 2026. Updated Sep 12, 2026. Spotted a problem? Tell us.
- Rethinking skills and prompts for GPT-6 AstraOpenAI · OpenAI Developers Blog · Sep 12, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- HeuristicDefine task completion clearly to avoid premature stoppingExplicitly define what 'done' means for a task to ensure AI models complete the full scope. · 10 min
- HeuristicShorten skill descriptions for better model context managementKeep skill descriptions brief and clear to help models pick the right skill and avoid context overload. · 15 min
- Anti-patternBuilding agent infrastructure from scratchAvoid piecing together agent infrastructure; use standardized SDKs and native sandbox support. · 5 min
- Setting tipAgent workspace portability with Manifest abstractionUse the Manifest abstraction to describe agent workspaces for portability across different providers. · 10 min
- ToolOpenAI Agents SDK for building agentsSDK for building agents that can inspect files, run commands, edit code, and work on long-horizon tasks. · 15 min
- Anti-patternOverly specific instructions for advanced AI modelsAvoid detailed itineraries; newer models handle nuance and ambiguity better. · 10 min
Questions people ask
- Which AI tools does this work with?
- Any current model. It was seen with ChatGPT in the sources.
- Where did this come from?
- OpenAI Developers Blog (OpenAI Developers Blog). Subvolts collected, normalized and credited it; the link above opens the original.