Unclear It is not clear what any of the three switches controls.
What
/claude-api is a built-in skill, a packaged set of instructions, that helps with code using the Claude API. It can now search your project for Claude model IDs that are deprecated or retired:
- Sonnet 4.5, Opus 4 and 4.1, Sonnet 4
- Haiku 3, Haiku 3.5
- Sonnet 3.5 and 3.7, Opus 3, Sonnet 3
- Claude 2
Two flags control this. A flag is a switch Anthropic's server can turn on remotely.
- The scan runs only when
tengu_cosmic_sketchis on, and it falls back to false. tengu_typed_trinketputs you in a test group: "control", "treatment" or "not_enrolled".- Only the "treatment" group gets a new section, "## Deprecated or Retired Models in This Project", added to the skill's instructions. The other groups only have the result recorded.
- The section tells Claude to mention the retiring model once. When you are not already running
/claude-api migrate, it also tells Claude to mention that subcommand. - The scan stops after 1000 ms by default, and its results are saved per project.
A third new flag, tengu_quiet_medlar, also appears, along with a helper that treats policySettings, userSettings and flagSettings differently. What these control was not traced.
For tengu_cosmic_sketch, the flag server returned off for this site's account and off for an anonymous baseline. No reading has been taken under this release yet. Nothing has been read about tengu_typed_trinket or tengu_quiet_medlar, so their state is unread, which is not the same as off.
Why
For most people this does nothing yet. If you are in the treatment group, Claude may point out old model IDs in your project and suggest /claude-api migrate.
tengu_cosmic_sketch Off by default, switched on for this accountThe shipped code defaults this off, and the flag server returned on for the one account this site reads on this version. That is the reading that makes the entry above worth a second look, and it still says nothing about your account.
This account: on · anonymous baseline: on · compiled default in v2.1.310: off
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.
tengu_typed_trinket 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 v2.1.310: 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 v2.1.310. It isn't a statement about your account. What a flag value here can and cannot tell you
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
It is not clear what any of the three switches controls.
New in this build: tengu_cosmic_sketch