Instead of guessing, or having the user investigate on their own, the AI agent should be able to tell them which files they would want to commit with the repo. Example:
i asked "which files does claude use to keep information about this repo that we should also commit along with the repo so that others have the same settings as we do locally when developing?"
claude said
Claude Code files & version control
In this repo right now
.claude/settings.local.json — the personal/local settings file. By convention it is not committed (machine-specific choices like approved permissions). Claude Code normally gitignores it automatically.
No CLAUDE.md and no .claude/settings.json yet.
Files meant to be committed (so collaborators share your setup)
File / dir
Purpose
Commit?
CLAUDE.md
Project instructions/memory Claude reads each session (conventions, how to run, gotchas)
✅ yes
.claude/settings.json
Shared project settings — permissions, hooks, env, model
✅ yes
.claude/settings.local.json
Personal overrides for just you
❌ no (gitignore)
.mcp.json
Shared MCP server config for the project
✅ yes (if used)
.claude/commands/
Custom slash commands
✅ yes
.claude/agents/
Custom subagents
✅ yes
.claude/hooks/
Hook scripts referenced by settings
✅ yes
Key distinction
.claude/settings.json → committed, team-wide
.claude/settings.local.json → yours, gitignored
Right now you only have the .local one, so nothing Claude-related is shared yet — and nothing Claude-related is tracked by git, nor ignored.
Recommended next steps
Add .claude/settings.local.json to .gitignore so personal settings aren't committed by accident.
Promote settings you want everyone to share into .claude/settings.json.
Add a CLAUDE.md documenting how to run/develop the app.
armed with that output the AI agent can actually do the work for us. and again we are trying to get the learner used to letting their ai modify files ....
Instead of guessing, or having the user investigate on their own, the AI agent should be able to tell them which files they would want to commit with the repo. Example:
i asked "which files does claude use to keep information about this repo that we should also commit along with the repo so that others have the same settings as we do locally when developing?"
claude said
# Claude Code files & version control
## In this repo right now
- `.claude/settings.local.json` — the **personal/local** settings file. By convention it is **not** committed (machine-specific choices like approved permissions). Claude Code normally gitignores it automatically.
- No `CLAUDE.md` and no `.claude/settings.json` yet.
## Files meant to be committed (so collaborators share your setup)
| File / dir | Purpose | Commit? |
|---|---|---|
| `CLAUDE.md` | Project instructions/memory Claude reads each session (conventions, how to run, gotchas) | ✅ yes |
| `.claude/settings.json` | **Shared** project settings — permissions, hooks, env, model | ✅ yes |
| `.claude/settings.local.json` | **Personal** overrides for just you | ❌ no (gitignore) |
| `.mcp.json` | Shared MCP server config for the project | ✅ yes (if used) |
| `.claude/commands/` | Custom slash commands | ✅ yes |
| `.claude/agents/` | Custom subagents | ✅ yes |
| `.claude/hooks/` | Hook scripts referenced by settings | ✅ yes |
## Key distinction
- `.claude/settings.json` → committed, team-wide
- `.claude/settings.local.json` → yours, gitignored
Right now you only have the `.local` one, so nothing Claude-related is shared yet — and nothing Claude-related is tracked by git, nor ignored.
## Recommended next steps
1. Add `.claude/settings.local.json` to `.gitignore` so personal settings aren't committed by accident.
2. Promote settings you want everyone to share into `.claude/settings.json`.
3. Add a `CLAUDE.md` documenting how to run/develop the app.
armed with that output the AI agent can actually do the work for us. and again we are trying to get the learner used to letting their ai modify files ....
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Instead of guessing, or having the user investigate on their own, the AI agent should be able to tell them which files they would want to commit with the repo. Example:
i asked "which files does claude use to keep information about this repo that we should also commit along with the repo so that others have the same settings as we do locally when developing?"
claude said
Claude Code files & version control
In this repo right now
.claude/settings.local.json— the personal/local settings file. By convention it is not committed (machine-specific choices like approved permissions). Claude Code normally gitignores it automatically.CLAUDE.mdand no.claude/settings.jsonyet.Files meant to be committed (so collaborators share your setup)
CLAUDE.md.claude/settings.json.claude/settings.local.json.mcp.json.claude/commands/.claude/agents/.claude/hooks/Key distinction
.claude/settings.json→ committed, team-wide.claude/settings.local.json→ yours, gitignoredRight now you only have the
.localone, so nothing Claude-related is shared yet — and nothing Claude-related is tracked by git, nor ignored.Recommended next steps
.claude/settings.local.jsonto.gitignoreso personal settings aren't committed by accident..claude/settings.json.CLAUDE.mddocumenting how to run/develop the app.armed with that output the AI agent can actually do the work for us. and again we are trying to get the learner used to letting their ai modify files ....