Source Intelligence
Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All of v2.1.39 Home All releases ←olderv2.1.38 v2.1.40newer→
Claude Code v2.1.39

Skill Improvement System

What

Claude Code now detects preferences and improvements you mention during skill execution and can automatically update skill definitions to remember them for future runs.

Usage

When running a skill, if you make requests like:

  • "Can you also ask me about X?"
  • "Please use a more casual tone"
  • "Don't do Z next time"

Claude Code will offer to apply these improvements to the skill definition file.

Details
  • Analyzes conversation for generalizable preferences vs one-time requests
  • Shows a prompt asking if you want to apply detected improvements
  • Updates the SKILL.md file in ~/.claude/skills/<skill-name>/
  • Displays "Skill [name] updated with improvements" when applied
Evidence

Skill improvement detection and application (search for "Skill improvement", "tengu_skill_improvement_detected", "updated with improvements")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

See this entry in the whole of v2.1.39 →