Anthropic Just Revealed 12 New Rules for Prompting Opus 5.5

Summary
The presenter from RoboNuggets reviews Anthropic’s official documentation and prompt engineering guide for the newly released Claude Opus 5.5. He outlines 12 specific tips and behavioral changes to optimize latency, cost, and task performance across coding, visual inputs, and multi-turn workflows.

What is shown

• [00:02] Anthropic documentation page: "Prompting Claude Opus 5.5".
• [00:23] Calibration of the effort level setting from "low" to "max", showing "medium" as the recommended default.
• [01:09] A testing prompt designed to run an identical user task across different effort levels to compare output and cost side-by-side.
• [01:29] Claude Code integration showing support for AGENTS.md (from version 2.1.277) to share instructions across coding agents.
• [02:07] Explanation of prompt caching preservation when using per-message effort changes mid-conversation.
• [02:46] The Claude web UI settings page showing the "Resets" box under Usage, including a free usage reset expiring October 23.
• [03:10] Error output demonstration showing a refusal triggered by asking Claude to show its reasoning steps (Details: [reasoning_extraction]).
• [03:45] System prompt instruction examples to prevent Opus 5.5 from unnecessarily re-evaluating settled answers in multi-turn conversations.
• [04:14] Checklist harness pattern for long-running agent tasks to avoid premature termination upon conversational end_turn.
• [04:51] Time budget pacing and using the phrase "Time matters" to accelerate agent completions.
• [05:33] Default frontend styling tendencies (cream backgrounds, italicized words) and feeding a design system to override them.
• [06:06] Tool enablement of Python libraries (PIL, OpenCV) for autonomous cropping and zooming into high-resolution technical drawings.

Claims & numbers

• The presenter states that Anthropic published an official prompt engineering guide specifically for Claude Opus 5.5.
• On Opus 5.5, the recommended default effort level is set to "medium", whereas Claude Opus 5 defaulted to "high".
• In Anthropic’s testing, "medium" effort on Opus 5.5 matches or exceeds Claude Opus 5 at "high" on coding and knowledge-work evaluations at lower cost.
• Claude Code version 2.1.277 added support to check for and load AGENTS.md if CLAUDE.md is absent.
• The free usage reset granted with the Opus 5.5 release expires on October 23.
• Prompts explicitly asking Claude Opus 5.5 to output its internal reasoning steps are now declined under the API category reasoning_extraction.
• Supplying specific time boundaries or the prompt instruction "Time matters: do not spend time that can be avoided, and the earlier a correct result is obtained, the better" measurably reduced completion times in multi-agent benchmarks.

Notable quotes

• [00:08] "What worked on previous models is now either costing you more or slowing you down."
• [03:26] "...the refusal reason simply states as reasoning extraction."
• [05:19] "Time matters here: do not spend time that can be avoided, and the earlier a correct result is obtained, the better."

Assessment
This is an educational explainer and practical review breaking down Anthropic’s official Claude Opus 5.5 prompt engineering documentation. The examples and UI interactions accurately reflect Anthropic’s released documentation, API settings, and model behavior guidelines.

Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.