OpenAI explains how to rethink skills, prompts and instructions for GPT-6 Astra
solyarisoftware · x · 2026-09-13
- OpenAI published guidance on redesigning skills, prompts and instructions for more capable models like GPT-6 (codename Astra).
- Key point: with stronger models, excessive rules, procedures and detailed instructions can backfire — they limit model autonomy and bloat context. OpenAI suggests shorter, more focused skills, progressive disclosure (load info only when needed), avoiding rigid procedures, pruning stale rules from AGENTS.md, and dropping manual checks the model already performs.
- Commenter Alessio Pomaro adds a distinction: the issue isn't precision in instructions but precision turning into prescriptiveness. Be maximally precise on goals, constraints and success criteria — but leave the path free when the path itself isn't a requirement.
Related event: OpenAI ships GPT-6 Astra guide: rethink skills, AGENTS.md and task prompts(5 posts)→
More from coding & agent
- Supercharging LLM classification with logprobs: soft classifiers you can threshold-tune — JnBrymn · 2026-10-28
- Codex daily build series day 23: skipping iCloud plugins for a reset — dotey · 2026-10-28
- Cloudflare Access adds fully customizable pages with HTML + Liquid templates — dinasaur_404 · 2026-10-28
- Opus retried every session 20.6 million times during a memory leak hunt — johnlindquist · 2026-10-28
- Open-Source Skill Teaches AI Agents to Design and Produce Videos — songguoxiansen · 2026-10-28
- Not Everything Needs an Agent: Fixed Steps Are Recipes, Save Agents for Judgment — adamse · 2026-10-28