Unclear It is not clear whether the result is still reported to Anthropic when the threshold is 0 and the check is off.
The propose_skills tool now compares a proposed SKILL.md with your current one before accepting it. It is the tool Claude uses to suggest a new or improved skill, which is a saved set of instructions Claude Code can follow for a task. Until now the tool only checked that the skill had been read first.
The new check works like this:
- Claude Code counts how many non-blank lines of your current SKILL.md the proposal keeps.
- It compares that count against a threshold. A valid threshold is from 50 to 100, and 0 turns the check off.
- If the proposal falls short, it is refused with the same message Claude Code gives when a skill was not read or has gone out of date.
The threshold comes from Anthropic's servers. The value built into the code is 0, so the check is off unless a different value is sent. Claude Code also reports the result of the check to Anthropic, both when it validates a proposal and when it runs the tool. That report includes whether the proposal would have passed at thresholds of 100, 90 and 50.
The check stops a proposed rewrite from quietly throwing away most of a skill you already have.
tengu_glittery_fern Not enough to sayNothing here resolved what this flag was doing on this version, so nothing here should be read as on or off.
This account: no value returned · anonymous baseline: no value returned · compiled default in v2.1.290: not a boolean we can read
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
Read once, for one account on one subscription tier, against v2.1.290. It isn't a statement about your account. What a flag value here can and cannot tell you
It is not clear whether the result is still reported to Anthropic when the threshold is 0 and the check is off.
New in this build: tengu_glittery_fern