One prompt line stops Claude from rewriting whole files for tiny edits
alex_verem · x · 2026-09-14
The post links to the official Claude Platform docs page "Prompting Claude Fable 5.1", summarizing behavioral differences and prompting patterns:
- File edits: The model prefers rewriting an entire file rather than making a small targeted edit — the result is identical but you pay for every token of the rewrite; one sentence in your prompt fixes this.
- Other behaviors: choosing the right effort level for latency/cost, requesting user-facing progress updates between tool calls, batching independent tool calls, and keeping conversation history append-only.
- Writing & formatting: default prose runs long and dense; density can be tuned via prompting.
- More: specifying what compaction summaries preserve, scoping changes/tests to the request, quoting retrieved sources properly, task completion, subagents, and vision.
Related event: Official docs offer prompt fix for Claude's whole-file rewrites(2 posts)→
More from coding & agent
- Agent audit metric: can you name which tests ran and files changed without reading its summary? — ronbodkin · 2026-10-31
- Dev: coding agent notifications only worth pushing when waiting on approval — petrusenko_max · 2026-10-31
- Unofficial OpenTargets MCP Server Brings Gene-Drug-Disease Data to LLMs — modelcontextprotocol · 2026-10-31
- zero-core-budget MCP Connector Adds Spending and ROI Tracking for Agents — modelcontextprotocol · 2026-10-31
- Designer uses Claude with subagents on screencap+audio to make the whole world drivable — tomjohndesign · 2026-10-31
- Reddit Thread Collects the Worst Things Coding Agents Did in Real Repos — Practical-Craft4967 · 2026-10-31