OpenAI recommends that developers moving to GPT-6 Astra cut back on accumulated instructions, blanket document-reading requirements and rigid approval rules, according to a blog post by Eric Provencher. The company says more capable models need less hand-holding, and that instructions piled up over time can eat up context or cause Astra to stop work too early. For teams that maintain agent workflows, this changes what a prompt file should contain.
What OpenAI recommends changing
Provencher's advice covers three areas: skills, AGENTS. md rules and task prompts. Skills are prompts stored as Markdown files that can include resources and scripts; their names and descriptions go into the model's context so Codex can pick the right skill for a given task. AGENTS. md holds the rules that govern repository work. Task prompts define what the agent should do and when the job is done. OpenAI recommends reviewing all three whenever a team switches models, because what worked for earlier models may already be too restrictive for Astra.
Too many skills force Codex to truncate descriptions, stripping out information it needs to choose correctly. Conflicting descriptions or overly broad scope also load instructions the model does not need. Provencher recommends keeping scope descriptions short and precise: a skill for Postgres schema migrations should fire only when creating or modifying a migration or when checking its rollout. If a skill covers multiple workflows, its main document should briefly point to the right supplementary docs and scripts, so the model reads only what it actually needs. Every extra read eats context and pushes the model closer to summarization.
Detailed step-by-step sequences can slow down newer models, since they handle nuance and ambiguity better on their own, according to Provencher. The same applies to repository rules. Requiring the model to read multiple documents or a full project overview before every change is overkill for a typo fix. Instead of forcing it to read architecture. md, database. md and deployment. md every time, Provencher suggests pointing to those docs selectively: architecture information when working on service boundaries, database docs when changing schemas, deployment notes when shipping. The docs also need to stay current.
What this means for teams running agents
For companies that have built agent workflows over the past year, the practical consequence is that prompt files and repository rules need periodic maintenance, not one-time setup. A small team with a handful of skills can review them in an afternoon; a larger organization with shared skills across many contributors faces a harder problem, because shared skills apply to every contributor's agents. What works for Sol or Luna might already be too restrictive for someone running Astra, so teams need to decide whether to keep one shared set or split it by model.
Explicit permissions can cut down on repeated confirmation requests for safe operations. For local tests using throwaway data with no production access, AGENTS. md can explicitly allow the agent to run tests, fix errors caused by the requested change and re-run affected tests without asking again. If earlier models went rogue and a team locked things down with strict approval rules, Provencher says it is time to revisit those rules when switching to Astra. OpenAI credits the model with better judgment, but it can also interpret old restrictions so literally that it stops even when the developer wants it to keep going.
Even without restrictions, Astra may stop earlier than GPT-5.6 Sol, Provencher notes. He recommends defining upfront what "done" means: if the agent should implement something, run it, check the results and fix errors, all of that needs to be in the prompt. A requirement to check in after the first implementation sets an earlier stopping point. These skill and project instruction recommendations build on OpenAI's recently published prompting tips for GPT-6 Astra, so the direction is consistent across the company's guidance.
The marker to watch is whether teams actually rewrite their AGENTS. md and skill files after switching models, and whether OpenAI publishes comparable guidance for its next model release. If the advice holds, the visible sign will be leaner prompt files in public repositories and fewer blanket reading requirements in shared agent setups. For business buyers of agent tooling, that is the point to check with vendors: whether their default configurations were written for the previous generation of models or reviewed for Astra.
