claude-use 0.4.0 → 0.6.0

package/README.md

@@ -74,6 +74,7 @@ Select an identity with:
 - `claude @<name>` — equivalent, once `claude-use shim enable` has been run (see [Install](#install))
 - `CLAUDE_ACCOUNT=<name> claude` — equivalent, via environment variable (this is `claude-use`'s own variable, read by its launcher; Anthropic's own multi-account convention is a plain `CLAUDE_CONFIG_DIR=<path> claude`, which `claude-use` builds on top of rather than replaces) — also needs the shim enabled first
 - `claude-use identity use <name>` — persistently, until changed again
+- `claude-use @<name>` — the same, terser: shorthand for `claude-use identity use <name>`, matching the `@name` convention the other forms above already use. Deliberately requires the `@` prefix and requires `@<name>` to be the *only* argument — identity names are user-chosen and unconstrained against `claude-use`'s own subcommand vocabulary (`identity`, `profile`, `rules`, `check`, `configure`, `doctor`, `shim`, `run`), so a bare `claude-use <name>` (no `@`) is deliberately left alone as an "unknown command" error rather than risking a future identity name colliding with a future subcommand name
 
 A directory rule (see below) can also pin a specific identity to a path, overriding whichever one is otherwise active — useful as a safety net so a particular client's directory always uses the right login regardless of habit.
 
@@ -213,6 +214,12 @@ Because the farm's content now depends on **(identity, resolved configuration pr
 
 Running two or more sessions concurrently under one identity — two terminals, each in a different client directory, is exactly the pattern directory rules are meant to support — means two resyncs can race to mutate the same shared farm toward two different resolved states. The launcher serialises this with a per-identity lock file (held for the duration of the resync, released before spawning `claude`) and builds each resync's changes as a scratch tree swapped into place with an atomic rename rather than mutating the live farm path-by-path in place, so a sibling session never observes a half-updated farm partway through someone else's resync.
 
+### Resolving a retained superseded farm
+
+Swapping in a resynced farm carries the identity's own real local data (credentials, `identity.json`, daemon/runtime state — anything that isn't a symlink or a directory the previous resync itself materialised) across from the superseded farm into the new one. When a top-level name exists in both, the swap never overwrites in either direction — it leaves the superseded farm on disk and reports it (`FARM_PREVIOUS_RETAINED`, or `FARM_SWAP_RECOVERED` when a crash-recovery pass on a later launch rediscovers it) rather than guessing which copy is more important. This can only happen when the underlying data genuinely differs in a way the tool has no way to judge safely on its own — the category system only tracks whether data is *shared across identities*, not whether it's *precious vs. disposable*.
+
+`claude-use identity resolve <name>` walks every retained `.{name}.previous.*` directory for that identity and resolves each collision interactively: keep the current farm's copy, keep the superseded farm's copy, or skip it for now (leaving it exactly as-is for a later run to ask about again). A superseded directory is only removed once every one of its own conflicts has been decided; skipping even one leaves the whole directory retained.
+
 ## Portable config: `.claude-use.json`
 
 `~/.claude-use/directory-rules.json` is local to one machine and keyed by absolute path — it doesn't survive being shared with a teammate, or even the same person cloning a repo to a different location. A `.claude-use.json` file committed at a project's root closes that gap. It's discovered exactly the way nested `CLAUDE.md` files are: every `.claude-use.json` found while walking upward from `$PWD` is collected, sorted shallowest-first, and folded into the cascade like a directory rule — except its scope is implicit (wherever the file lives, and everything below it) rather than an explicit `path` field, so it works identically no matter where the repo is checked out.
@@ -277,7 +284,7 @@ claude-use identity set <name> --allow-ambient-credential   # persistently, for
 
 | What you're setting | Global (persistent) | Temporary (this run only) | Directory-scoped (persistent) |
 |---|---|---|---|
-| **Identity** | `claude-use identity use <name>` (writes `~/.claude-use/active-identity`) | `claude-use run @<name>` / `claude @<name>` (needs `claude-use shim enable`) / `CLAUDE_ACCOUNT=<name> claude` (same) | `claude-use rules add <path> --identity <name>`; or `.claude-use.json`'s `"identity"` |
+| **Identity** | `claude-use identity use <name>` / `claude-use @<name>` (writes `~/.claude-use/active-identity`) | `claude-use run @<name>` / `claude @<name>` (needs `claude-use shim enable`) / `CLAUDE_ACCOUNT=<name> claude` (same) | `claude-use rules add <path> --identity <name>`; or `.claude-use.json`'s `"identity"` |
 | **Configuration profile** | `claude-use profile set-default <name>`; or `claude-use identity set-default-profile <identity> <profile>` | `claude --config-profile <name>` / `CLAUDE_USE_CONFIG_PROFILE=<name> claude` | `claude-use rules add <path> --profile <name>`; or `.claude-use.json`'s `"configProfile"` |
 | **A category** | `claude-use profile set <name> --category history=true`; or `claude-use configure <identity>` | `claude --category history=true[,knowledge=false,...]` / `CLAUDE_USE_CATEGORY_OVERRIDE="history=true,knowledge=false"` | `claude-use configure <identity>` run from inside the ruled directory; or `.claude-use.json`'s `"categories"` |
 | **An individual entry** | `claude-use profile set <name> --entry "path"=true`; or `claude-use configure <identity> <path>` | `claude --share <path>[,<path>,...]` / `claude --hide <path>[,<path>,...]` / `CLAUDE_USE_ENTRY_OVERRIDE="path=true,otherpath=false"` | `claude-use configure <identity> <path>` run from inside the ruled directory; or `.claude-use.json`'s `"entries"` |
@@ -291,9 +298,11 @@ The scriptable `claude-use profile set ...` commands exist alongside the interac
 ```
 claude-use identity add <name>
 claude-use identity use <name>
+claude-use @<name>                          # shorthand for `identity use <name>`
 claude-use identity list
 claude-use identity set-default-profile <identity> <profile>
 claude-use identity set <name> [--allow-ambient-credential | --no-allow-ambient-credential]
+claude-use identity resolve <name>          # interactively resolve a retained superseded farm's conflicts
 
 claude-use profile create <name> [--extends <name>,<name>,...]
 claude-use profile list