CLAUDE.md rewritten from Anthropic's official Opus 5.5 prompting guide — fewer tokens, medium effort default
Puzzled-Ad-6854 · reddit · 2026-09-23
The author dug through Anthropic's official Opus 5.5 prompting guide and rebuilt their CLAUDE.md accordingly. Key doc findings: 5.5 explains its work plainly and finishes tasks with fewer tokens; medium-effort 5.5 matched or beat high-effort Opus 5 on Anthropic's coding and knowledge-work evals; medium is now the default effort, and the same setting makes 5.5 think more per turn at higher levels.
Anthropic's context engineering article also revealed they cut over 80% of Claude Code's system prompt with no measurable coding eval loss. The author's updated config keeps preferences for concise communication, task scope, approval for destructive actions and a Context7 doc-lookup rule, while dropping hard sentence limits and repeated instructions. File and setup notes are on GitHub, with a caveat that some rules are personal preferences.
More from coding & agent
- Pigment Agents now reads and posts in Slack, inheriting user permissions — mattturck · 2026-11-12
- Map of the agent economy: 50+ startups across payments, spend control, identity, memory — venturetwins · 2026-11-12
- OpenCode launches Browse, an on-device browser agent with local logins — Saul_Loveman · 2026-11-12
- VentureTwins maps the B2A economy: dozens of agent infra and app companies in first series installment — venturetwins · 2026-11-12
- Google releases 76-page 'Agents Companion' whitepaper on agent engineering — mdancho84 · 2026-11-12
- Spinning up containers for 20-line agent tools is backwards: Safescript compiles provable, 50ms-deploy apps — uriwa · 2026-11-12