Anthropic says Claude Code can drop 80% of its system prompt in Claude 5
AI寒武纪 · wechat · 2026-07-25
Anthropic says Claude Code’s context-engineering playbook has changed dramatically for the Claude 5 generation.
- The company says it removed more than 80% of the old system prompt without measurable coding-quality loss.
- The new guidance emphasizes letting the model infer intent, designing tools and interfaces well, using progressive disclosure, and relying less on hand-written repetition.
- Memory is now handled automatically, while Skills, CLAUDE.md, artifacts, rubric files, and code-based references are treated as layered context sources.
- Anthropic also introduced claudedoctor, a command meant to help users slim down Skills and CLAUDE.md files and adopt the new workflow.
The post frames these changes as a broader shift from rigid instructions to richer, more adaptive context design for agentic coding.
Related event: Anthropic Slashes Claude Code System Prompts by Over 80%(11 posts)→
More from coding & agent
- The browser main thread is expensive: a practical guide to JavaScript and CSS animation cost — jh3yy · 2026-09-11
- Inspired by OpenAI's 10,000-agent run, dev open-sources a crowdsourced agent problem-solving platform — Benjaminsen · 2026-09-11
- Lucid: open-source Mac app keeps your laptop awake only while AI agents run — Pitiful_Hedgehog_600 · 2026-09-11
- banteg's snail project crowdsources AI agents to finish matching Snail Mail's 20 remaining functions — banteg · 2026-09-11
- Alex Townsend posts 200 open problems in numerical linear algebra for humans and AI agents — IgorCarron · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11