change-core-self — Claude Code command: what it does, how to install
change-core-self asks you about the project, works out a fitting identity for the agent from scratch, and stores it as a rule file loaded on every session.
no license ★ 406 updated 2026-06-25
.claude/commands/change-core-self.mdWhat change-core-self does
Run it with no arguments and the questions start: what the artifact really is, what counts as success, how long the code lives, who else will read it, which constraints are hard. They arrive in batches through AskUserQuestion. Then comes the reasoning, spelled out in the open: KISS, YAGNI, DRY, modularity, clean architecture, file-size caps, tests — each one gets kept, dropped or flipped for this repo, with the tradeoff written next to it.
The result lands in .claude/rules/atlas-self-profile.md. It works as an overlay and outranks misc/self/engineering.md plus the posture sections of misc/self/atlas.md, so the stock principles survive only where the new file adopts them again. You see the derived identity and approve it before anything is saved. Two short modes skip the interview: show prints the active profile, reset deletes it. The model will not fire this one on its own; you type it.
Every default can be thrown out at once, and for a shader toy or a game-jam entry that may be the right call. Expect the opposite if you want a tidy layered codebase: the command bends the standard to fit the project, not the project to fit the standard.
What it helps with
- rewrite the agent's standing instructions for one specific repo
- get the person paying to define what good looks like
- drop DRY and modularity where they fight the work
- check which profile is currently loaded
- restore the stock posture with a single call
Example prompts for the agent
/change-core-self— interview me about this project and write the profile/change-core-self show— print the active profile/change-core-self reset— delete the overlay and go back to defaults
How to install change-core-self
Manually: put the files in place
.claude/commands/change-core-self.mdCall it with /change-core-self. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/change-core-self.mdCall it with /prompts:change-core-self. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/change-core-self.prompt.mdManually: put the files in place
.gemini/commands/change-core-self.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about change-core-self
How do I install change-core-self in Claude Code?
Place the command file at .claude/commands/change-core-self.md inside your project. It then runs as /change-core-self.
What does change-core-self do?
It interviews you about the project, judges each engineering default against it, and saves the resulting agent identity to .claude/rules/atlas-self-profile.md.
Do KISS, YAGNI and DRY still apply afterwards?
Only where the generated profile re-adopts them, since the overlay takes precedence over misc/self/engineering.md and the posture parts of misc/self/atlas.md. reset removes the file and the defaults come back.