agent-toggle 0.8.1__tar.gz → 0.10.0__tar.gz

{agent_toggle-0.8.1 → agent_toggle-0.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agent-toggle
-Version: 0.8.1
+Version: 0.10.0
 Summary: Temporarily disable / re-enable AI-agent skills, agents, commands, rules, plugins and MCP servers across harnesses (Claude Code, Codex, Grok, OpenCode, OpenClaw, Copilot, Vibe, Devin, Antigravity)
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/weskao/agent-toggle
@@ -32,6 +32,8 @@ Dynamic: license-file
 
 # agent-toggle
 
+PyPI: <https://pypi.org/project/agent-toggle/>
+
 Temporarily disable and restore AI-agent resources — skills, agents, commands,
 rules, plugins (mods included), MCP servers — across Claude Code, Codex, Grok CLI, OpenCode,
 OpenClaw, Copilot, Vibe, Devin and Antigravity — and show what each one costs at
@@ -40,6 +42,27 @@ session start.
 **Nothing is ever deleted.** Everything is parked and recorded, and `enable`
 puts it back where it came from.
 
+## At a glance
+
+<!-- Absolute URLs so the images also render on PyPI, which uses this README. -->
+
+**`agent-toggle ui`**: one list per harness, with what each resource costs in tokens.
+Toggle rows, see the token change before you apply it, and apply it as one batch that
+`undo` reverses ([Interactive picker](#interactive-picker)).
+
+![agent-toggle ui: the interactive picker](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/ui-picker.jpeg)
+
+**`agent-toggle config`**: every setting in one menu. Turn harnesses on or off and set the
+language, update check and Telegram notifications. It also runs tools such as `doctor` and
+`undo` ([Settings & config menu](#settings--config-menu)).
+
+![agent-toggle config: the settings menu](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/config-menu.jpeg)
+
+**`/agent-toggle` inside Claude Code**: the same picker as a pane, for turning skills, MCP
+servers, plugins and mods on or off without leaving the session ([Claude Code pane mod](#claude-code-pane-mod)).
+
+![/agent-toggle: the picker as a Claude Code pane](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/claude-pane-mod.jpeg)
+
 ## Why
 
 | target | native mechanism | the gap |
@@ -59,18 +82,19 @@ uv tool install agent-toggle
 agent-toggle --version
 ```
 
-Straight from this repository, pinned to a release tag (`v0.5.1` is the current one;
-an unpinned `git+` URL installs whatever is on the default branch):
+Straight from this repository, pinned to a release tag (replace `vX.Y.Z` with one from
+[tags](https://github.com/weskao/agent-toggle/tags); an unpinned `git+` URL installs
+whatever is on the default branch):
 
 ```sh
-uv tool install git+https://github.com/weskao/agent-toggle@v0.5.1
+uv tool install git+https://github.com/weskao/agent-toggle@vX.Y.Z
 agent-toggle --version
 ```
 
 Alternatively, use `pipx`:
 
 ```sh
-pipx install agent-toggle             # from PyPI; or: pipx install git+https://github.com/weskao/agent-toggle@v0.5.1
+pipx install agent-toggle             # from PyPI; or: pipx install git+https://github.com/weskao/agent-toggle@vX.Y.Z
 ```
 
 From a checkout, either install it editable or run it in place:
@@ -86,12 +110,13 @@ Python 3.10+, no runtime dependencies. Then write the skill shim:
 
 ```sh
 agent-toggle install-shims             # installed, or: ./install.sh (a thin wrapper)
-python3 agent_toggle.py install-shims  # from a checkout, no install needed
+python3 agent_toggle.py install-shims  # from a checkout
 ```
 
 `install-shims` writes a thin skill shim (`skills/agent-toggle/SKILL.md`) into
 every installed harness that supports skills, so `/agent-toggle` works from any
-of them. It writes nothing else into a harness home: parked items live under
+of them. The shim calls `agent-toggle` on `PATH` and never a checkout path, so install
+the package (or `pip install -e .`) before relying on it. It writes nothing else into a harness home: parked items live under
 `~/.agent-toggle/parked/`, so a harness-home `.gitignore` needs no park-dir
 line. Older versions appended `<dir>-disabled/` lines there; inside a git work
 tree, `install-shims` reports such stale lines once `migrate` has emptied the
@@ -133,6 +158,18 @@ Without the extra everything else works exactly as before. Only `config test`,
 [settings menu](#settings--config-menu) need it; bare `config` still opens, and
 those rows read `install agent-toggle[telegram]`.
 
+**Optional: the Claude Code pane mod** (the picker as a pane inside Claude Code, claude harness
+only). It is not part of the PyPI package; with `agent-toggle` already on `PATH` from one of the
+installs above, type this at the prompt of a Claude Code terminal session (2.1.275 or later):
+
+```
+/plugin install agent-toggle --marketplace weskao/agent-toggle
+```
+
+Answer `y` to add the marketplace, pick a scope (user is the default), then run `/agent-toggle`.
+
+See [Claude Code pane mod](#claude-code-pane-mod).
+
 ## Shell completion
 
 Each release publishes static completion files for bash, zsh and fish as the
@@ -164,7 +201,7 @@ agent-toggle <command> [args]          # or: python3 agent_toggle.py <command> [
 | `disable\|enable all <name>...` | every type on the harness with that name at once (skill, command, MCP, ...); a name with no match is an `error` row; one batch, so `undo` reverses it |
 | `enable --all` | put back **every** disabled item (`--harness H` narrows it, `--project <dir>` takes only that project's) |
 | `undo` | reverse the last logged batch (`--dry-run` shows the plan) |
-| `profile save\|apply\|diff\|list` | named sets of live items; see [Profiles](#profiles) |
+| `profile save\|apply\|diff\|list\|delete\|rename` | named sets of live items; see [Profiles](#profiles) |
 | `config` | settings menu (curses, else a numbered list); `config --json` lists every setting and the package version; see [Settings & config menu](#settings--config-menu) |
 | `config test\|sync-ci` | send one Telegram test message / set the GitHub CI secrets; needs the `[telegram]` extra; see [CI notifications](#ci-notifications) |
 | `help [command]` | styled help; `help ui` = `--help ui` = `ui --help`; see [Help](#help) |
@@ -256,7 +293,7 @@ agent-toggle             # same, on a terminal; elsewhere prints help and exits
 ```
 
 ```
- agent-toggle  v0.5.1                                 5 resources · ~6.3k tok live · 1 staged (+300 tok)
+ agent-toggle  vX.Y.Z                                 5 resources · ~6.3k tok live · 1 staged (+300 tok)
  All 5 │ claude 4 │ codex 2
  🔍 Type to search (name, path, fuzzy)                                           type: All  sort: name
  Skills 3 ─────────────────────────────────────────────────────────────│ beta
@@ -297,7 +334,7 @@ What is on screen:
 
 | key | action |
 |---|---|
-| `Space` | toggle the highlighted row (live / parked) and stay on it (while filtering, `Space` types a space) |
+| `Space` | toggle the highlighted row (live / parked) and stay on it (while searching, the first `Space` separates search terms and a second one toggles) |
 | `Tab` | toggle the highlighted row and advance to the next |
 | `Enter` | apply every staged change (with `--dry-run`: show the plan) |
 | `Esc` / `Ctrl-C` | cancel; nothing is applied |
@@ -332,7 +369,8 @@ off under `--dry-run`. Esc closes the prompt.
 Nothing happens while the picker is open. Changes are staged, the screen is
 torn down, and only then do the real operations run, so their output (which
 companion files moved, which were kept because they are shared) is readable
-instead of fighting curses for the terminal.
+instead of fighting curses for the terminal. Claude Code sessions running the
+[pane mod](#claude-code-pane-mod) pick up the Claude changes within 5 s and reload them.
 
 Built on stdlib `curses`, so there is nothing to install on macOS and Linux
 (on Windows, `pip install "agent-toggle[windows]"` pulls `windows-curses`).
@@ -384,10 +422,10 @@ end of input, `q` or an empty line exits `0`
 
 | group | rows |
 |---|---|
-| General | Check for updates, Color, Language, Default harness |
+| General | Check for updates, Color, Logo, Language, Default harness |
 | Harnesses | one On/Off per harness, with `found ~/.x` or `not on this machine` |
 | Picker | default sort, harness filter, type filter |
-| Notifications | Telegram bot token (masked), Telegram chat ID |
+| Notifications | Telegram bot token (masked, last 4 characters shown), Telegram chat ID |
 | Tools | Health check (`doctor`; then asks y/n per fix it can run), Install shims, Undo last change (asks y/n), Send test message, Sync CI secrets (asks y/n); their output shows in an overlay |
 
 Keys: `↑` `↓` move (wraps, skips headings), `←` `→` cycle a value, `Enter` / `Space` change or run
@@ -403,6 +441,7 @@ tool does not know are kept):
 |---|---|---|---|
 | `update_check` | `true` / `false` | `true` | `AGENT_TOGGLE_UPDATE_CHECK` |
 | `color` | `auto` / `always` / `never` | `auto` | `AGENT_TOGGLE_COLOR` |
+| `logo` | `color` / `mono` / `animated` / `off` | `animated` | `AGENT_TOGGLE_LOGO` |
 | `language` | `en` / `zh-TW` | `en` | `AGENT_TOGGLE_LANG` |
 | `default_harness` | a harness name | `claude` | `AGENT_TOGGLE_DEFAULT_HARNESS` |
 | `harness.<name>` | `true` / `false`, for `claude` `codex` `grok` `opencode` `openclaw` `copilot` `vibe` `devin` `agy` | `true` | |
@@ -459,6 +498,13 @@ options and examples; `help help` is the overview, `help version` shows how to p
 and an unknown command exits `2`. It honours `--color` and `NO_COLOR`, and is
 translated when `language` is `zh-TW`.
 
+The overview opens with the agent-toggle logo (cyan to blue to magenta): the large one from 97
+columns, a compact one from 49, none below that. Piped output gets the plain glyphs, a stream
+that cannot encode them gets none. The `logo` setting picks `color`, `mono` (one colour),
+`animated` (one light sweep, only on a colour terminal) or `off`. The config menu shows the same
+logo, centred, when the window is tall enough to fit the whole menu beside it; there,
+`animated` sweeps on entry and again after every 5 s without a key, and any key stops it.
+
 ## What each harness supports
 
 | harness | home | skill | agent | command | rule | plugin | mcp |
@@ -555,7 +601,8 @@ visible rather than silent.
   hand-rolled serializer would mangle comments — a text slice is lossless and
   much less code.
 
-MCP changes need a **new session** to take effect.
+MCP changes need a **new session** to take effect, except in a Claude Code session running the
+agent-toggle mod, which runs `/mcp disable|enable <name>` for you (see the mod section).
 
 ### Claude MCP scopes
 
@@ -586,6 +633,8 @@ agent-toggle profile diff work                       # what apply would do; writ
 agent-toggle profile apply ~/dotfiles/agent-toggle/work.json --dry-run
 agent-toggle profile apply work
 agent-toggle profile list
+agent-toggle profile rename work flutter             # -> ~/.agent-toggle/profiles/flutter.json
+agent-toggle profile delete flutter
 ```
 
 **`apply` toggles only the items the profile mentions.** An item the profile
@@ -614,6 +663,14 @@ one batch, logged, undoable with `undo`).
   so a project profile can never disable the user's items.
 - `save` skips (with a warning) any item whose name `apply` would refuse, such as
   an MCP server called `team/search`, so a saved profile always applies.
+- `save` to an existing name overwrites it with the live state now; that is how a
+  profile is updated.
+- `delete <name>` and `rename <old> <new>` work on stored profiles only (a file
+  you saved with `--out` is yours to move or remove) and change no item. Both exit
+  `2` for a name with no stored profile or a path; `rename` also exits `2` when
+  `<new>` already exists (it never overwrites) or is not a valid profile name, and
+  keeps the file byte for byte. Neither takes `--dry-run` or `--project`: a
+  project profile is renamed or deleted by its name like any other.
 
 ## Undo and `enable --all`
 
@@ -793,6 +850,7 @@ bookkeeping.
 | `state.json` | current disabled list (schema v3; atomic write, mode `0600`), saved before each flag write or dir move with the op in flight under an optional `pending` key, and after each MCP / plugin item |
 | `lock` | held by `disable` / `enable` / `enable --all` / `undo` / `profile apply` / `migrate` / `ui` for the whole batch; a second run waits 5 s then exits `3` (stale after 10 min *and* its PID is gone; a live batch refreshes it per item; the error says when its PID is gone, i.e. a killed run left it) |
 | `log.jsonl` | one line per operation (mode `0600`), see below |
+| `reload.json` | the Claude rows the last `ui` Enter changed, rewritten each time; open Claude Code sessions running the mod poll it and reload them |
 | `mcp-backups/` | `<harness>__<server>.json`, or `<sha8>__<harness>__<server>.json` for a project `.mcp.json` (mode `0600` -- may hold auth headers) |
 | `companions/` | parked exclusive helper files |
 | `parked/user/<owner>/<dir>/` | user-scope items, e.g. `parked/user/claude/skills/<name>` (`<owner>` = the harness whose home holds the dir; a dir shared by several harnesses parks once, under its owner); an absolute OpenCode `skills.paths` dir parks under `parked/user/opencode/ext-<sha8>/` (`<sha8>` of its resolved path). Nothing is ever parked inside a harness home |
@@ -870,6 +928,69 @@ deletion lines in `git status`.
 unresolved paths makes every companion look like it lives outside the harness.
 Both sides are resolved first.
 
+## Claude Code pane mod
+
+A Claude Code mod in `mod/` that shows the picker as a pane inside Claude (not to be confused
+with the [Mods](#mods) group the picker lists). It is driven by the installed `agent-toggle`
+CLI (`cost --json`, `disable|enable --json`) and works for the claude harness only; the curses
+`agent-toggle ui` stays the cross-harness UI.
+
+Install it with `/plugin install agent-toggle --marketplace weskao/agent-toggle` (see
+[Install](#install)); `claude plugin update agent-toggle@agent-toggle` picks up a new release. To
+run your own checkout instead (edits reload as you save):
+
+```sh
+claude --plugin-dir ./mod    # then run /agent-toggle inside Claude
+```
+
+- `/agent-toggle` or `/agent-toggle open` opens the pane, `/agent-toggle close` closes it,
+  `/agent-toggle toggle` does whichever applies; any other argument that is not a CLI command (below) prints the usage. Ctrl-X X
+  also closes it. Claude can do the same when you ask it to open or close agent-toggle: the mod
+  registers a `pane` tool (`mcp__agent-toggle__pane`, `action`: `open` | `close` | `toggle`),
+  e.g. "open agent-toggle", "close the agent-toggle pane", "toggle the agent-toggle pane".
+  Turning resources on or off is unchanged: toggle a row in the pane, or ask Claude in plain
+  words ("disable the demo-skill skill", "add it back"), which runs the CLI through the skill shim.
+- It looks like the `ui` picker: a title with the live-token total, a search line, rows under
+  type headings with counts, green `●` live / yellow `○` parked, tokens and a bar per row (a
+  parked row shows `(N)`, what restoring it would load), the cursor row in reverse video, and a
+  key-chip footer. The list scrolls with the cursor; the status line shows its position (`7/23`),
+  and the first and last list lines count the rows out of view (`↑ 4`, `↓ 12`).
+- The mouse wheel over the list moves the cursor, as in the terminal picker; no click needed.
+- Click the list once so it takes keys (Esc hands the keyboard back to the prompt). Then `↑` `↓`
+  `PgUp` `PgDn` `Home` `End` (also `Ctrl-P` / `Ctrl-N`) move; `Space` or `Enter` toggles the row,
+  `Tab` toggles and moves on; `/` or any other letter starts a search (terms ANDed over
+  `type/name` and the group, letters-in-order if nothing matches; a second `Space` toggles);
+  `s` sorts by name or cost; `n` toggles every type with this row's name (skill, command, MCP,
+  ...), like `disable all <name>`; `Ctrl-U` clears the search. A click moves the cursor.
+- A toggle applies at once (no staging, no Enter to apply); the toast names the target and the state
+  with an emoji and a bracketed tag (no colour codes, so nothing garbles; on the Linux console and the
+  legacy Windows console, where emoji draw as boxes, the tag alone): `✅ [ENABLED] skill: foo`, `⛔ [DISABLED] mod: bar (/reload-plugins)`,
+  `⛔ [DISABLED] rule: baz (takes effect in a new session)`, `⛔ [DISABLED] all: foo` after `n`, or `❌ [FAILED] disable skill: foo - <error>`.
+  The row flips the moment you press (and flips back
+  if the CLI fails); the list reloads in the background. While a toggle runs, further keys are ignored.
+- Opening the pane shows the same braille spinner and `Loading resources…` as `agent-toggle ui`
+  until the list arrives, and a spinner beside `working…` while a toggle runs.
+- Not in the pane: harness tabs (claude only), and the `t` `p` `a` keys. Where Claude draws no
+  such pane (the mobile app, VS Code) it is a plain list of buttons instead.
+- `/agent-toggle <command> ...` with `disable`, `enable`,
+  `undo`, `list`, `status`, `cost` or `doctor` runs the CLI instead of opening the pane and prints
+  one line per result row: `/agent-toggle disable skill foo bar` (it adds `--json`, and
+  `--harness claude` unless you pass `--harness`). A toggle reloads an open pane.
+- Inside Claude Code the mod then applies the change to the running session by queueing the slash
+  command for you, once per kind: skills and commands `/reload-skills`, plugins (mods included)
+  `/reload-plugins`, an MCP server `/mcp disable|enable <name>`. The toast names it, e.g.
+  `⛔ [DISABLED] skill: foo (/reload-skills)`. Agents and rules have no reload and still say
+  `takes effect in a new session`. If Claude refuses a queued command, a toast says so.
+- `agent-toggle ui` in another terminal applies on Enter and writes the Claude rows it changed to
+  `~/.agent-toggle/reload.json`. Every open session running the mod checks that file's mtime every
+  5 s (one `stat`) and queues the same reloads, toasting `agent-toggle ui: /reload-skills`. A file
+  older than the session is ignored.
+- Needs `agent-toggle` on `PATH`.
+- On Windows `agent-toggle` must resolve to a real executable on `PATH` (a pip or uv install
+  provides `agent-toggle.exe`).
+
+The design is in `docs/MOD.md`.
+
 ## Design and roadmap
 
 Architecture, harness survey, cost model, known gaps and the phased roadmap
@@ -887,6 +1008,13 @@ ruff check .
 stdlib `unittest`, no fixtures, no network. Every test runs against a
 throwaway temp `HOME` and a stubbed `claude` CLI.
 
+The pane mod has its own checks (they need the `claude` CLI, so CI does not run them):
+
+```sh
+claude plugin validate mod
+claude plugin test mod
+```
+
 `python3 tools/wheel_smoke.py` builds the wheel, installs it without extras into a
 fresh venv (what a PyPI user gets) and runs the CLI against a temp `HOME`; CI runs
 it on macOS, Linux and Windows.
@@ -929,6 +1057,8 @@ prints a `::notice::` and exits 0, so it never adds a second red X.
 ## Releasing
 
 A release is a git tag `vX.Y.Z` (where `X.Y.Z` must match `agent_toggle.__version__`).
+Bump `version` in `mod/.claude-plugin/plugin.json` to the same value by hand; nothing checks it,
+and without it `claude plugin update` does not hand marketplace installs the new mod.
 Pushing the tag triggers `.github/workflows/release.yml`:
 
 1. **Build** job checks that the tag matches `__version__`, builds a wheel and

{agent_toggle-0.8.1 → agent_toggle-0.10.0}/README.md

@@ -1,5 +1,7 @@
 # agent-toggle
 
+PyPI: <https://pypi.org/project/agent-toggle/>
+
 Temporarily disable and restore AI-agent resources — skills, agents, commands,
 rules, plugins (mods included), MCP servers — across Claude Code, Codex, Grok CLI, OpenCode,
 OpenClaw, Copilot, Vibe, Devin and Antigravity — and show what each one costs at
@@ -8,6 +10,27 @@ session start.
 **Nothing is ever deleted.** Everything is parked and recorded, and `enable`
 puts it back where it came from.
 
+## At a glance
+
+<!-- Absolute URLs so the images also render on PyPI, which uses this README. -->
+
+**`agent-toggle ui`**: one list per harness, with what each resource costs in tokens.
+Toggle rows, see the token change before you apply it, and apply it as one batch that
+`undo` reverses ([Interactive picker](#interactive-picker)).
+
+![agent-toggle ui: the interactive picker](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/ui-picker.jpeg)
+
+**`agent-toggle config`**: every setting in one menu. Turn harnesses on or off and set the
+language, update check and Telegram notifications. It also runs tools such as `doctor` and
+`undo` ([Settings & config menu](#settings--config-menu)).
+
+![agent-toggle config: the settings menu](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/config-menu.jpeg)
+
+**`/agent-toggle` inside Claude Code**: the same picker as a pane, for turning skills, MCP
+servers, plugins and mods on or off without leaving the session ([Claude Code pane mod](#claude-code-pane-mod)).
+
+![/agent-toggle: the picker as a Claude Code pane](https://raw.githubusercontent.com/weskao/agent-toggle/main/docs/images/claude-pane-mod.jpeg)
+
 ## Why
 
 | target | native mechanism | the gap |
@@ -27,18 +50,19 @@ uv tool install agent-toggle
 agent-toggle --version
 ```
 
-Straight from this repository, pinned to a release tag (`v0.5.1` is the current one;
-an unpinned `git+` URL installs whatever is on the default branch):
+Straight from this repository, pinned to a release tag (replace `vX.Y.Z` with one from
+[tags](https://github.com/weskao/agent-toggle/tags); an unpinned `git+` URL installs
+whatever is on the default branch):
 
 ```sh
-uv tool install git+https://github.com/weskao/agent-toggle@v0.5.1
+uv tool install git+https://github.com/weskao/agent-toggle@vX.Y.Z
 agent-toggle --version
 ```
 
 Alternatively, use `pipx`:
 
 ```sh
-pipx install agent-toggle             # from PyPI; or: pipx install git+https://github.com/weskao/agent-toggle@v0.5.1
+pipx install agent-toggle             # from PyPI; or: pipx install git+https://github.com/weskao/agent-toggle@vX.Y.Z
 ```
 
 From a checkout, either install it editable or run it in place:
@@ -54,12 +78,13 @@ Python 3.10+, no runtime dependencies. Then write the skill shim:
 
 ```sh
 agent-toggle install-shims             # installed, or: ./install.sh (a thin wrapper)
-python3 agent_toggle.py install-shims  # from a checkout, no install needed
+python3 agent_toggle.py install-shims  # from a checkout
 ```
 
 `install-shims` writes a thin skill shim (`skills/agent-toggle/SKILL.md`) into
 every installed harness that supports skills, so `/agent-toggle` works from any
-of them. It writes nothing else into a harness home: parked items live under
+of them. The shim calls `agent-toggle` on `PATH` and never a checkout path, so install
+the package (or `pip install -e .`) before relying on it. It writes nothing else into a harness home: parked items live under
 `~/.agent-toggle/parked/`, so a harness-home `.gitignore` needs no park-dir
 line. Older versions appended `<dir>-disabled/` lines there; inside a git work
 tree, `install-shims` reports such stale lines once `migrate` has emptied the
@@ -101,6 +126,18 @@ Without the extra everything else works exactly as before. Only `config test`,
 [settings menu](#settings--config-menu) need it; bare `config` still opens, and
 those rows read `install agent-toggle[telegram]`.
 
+**Optional: the Claude Code pane mod** (the picker as a pane inside Claude Code, claude harness
+only). It is not part of the PyPI package; with `agent-toggle` already on `PATH` from one of the
+installs above, type this at the prompt of a Claude Code terminal session (2.1.275 or later):
+
+```
+/plugin install agent-toggle --marketplace weskao/agent-toggle
+```
+
+Answer `y` to add the marketplace, pick a scope (user is the default), then run `/agent-toggle`.
+
+See [Claude Code pane mod](#claude-code-pane-mod).
+
 ## Shell completion
 
 Each release publishes static completion files for bash, zsh and fish as the
@@ -132,7 +169,7 @@ agent-toggle <command> [args]          # or: python3 agent_toggle.py <command> [
 | `disable\|enable all <name>...` | every type on the harness with that name at once (skill, command, MCP, ...); a name with no match is an `error` row; one batch, so `undo` reverses it |
 | `enable --all` | put back **every** disabled item (`--harness H` narrows it, `--project <dir>` takes only that project's) |
 | `undo` | reverse the last logged batch (`--dry-run` shows the plan) |
-| `profile save\|apply\|diff\|list` | named sets of live items; see [Profiles](#profiles) |
+| `profile save\|apply\|diff\|list\|delete\|rename` | named sets of live items; see [Profiles](#profiles) |
 | `config` | settings menu (curses, else a numbered list); `config --json` lists every setting and the package version; see [Settings & config menu](#settings--config-menu) |
 | `config test\|sync-ci` | send one Telegram test message / set the GitHub CI secrets; needs the `[telegram]` extra; see [CI notifications](#ci-notifications) |
 | `help [command]` | styled help; `help ui` = `--help ui` = `ui --help`; see [Help](#help) |
@@ -224,7 +261,7 @@ agent-toggle             # same, on a terminal; elsewhere prints help and exits
 ```
 
 ```
- agent-toggle  v0.5.1                                 5 resources · ~6.3k tok live · 1 staged (+300 tok)
+ agent-toggle  vX.Y.Z                                 5 resources · ~6.3k tok live · 1 staged (+300 tok)
  All 5 │ claude 4 │ codex 2
  🔍 Type to search (name, path, fuzzy)                                           type: All  sort: name
  Skills 3 ─────────────────────────────────────────────────────────────│ beta
@@ -265,7 +302,7 @@ What is on screen:
 
 | key | action |
 |---|---|
-| `Space` | toggle the highlighted row (live / parked) and stay on it (while filtering, `Space` types a space) |
+| `Space` | toggle the highlighted row (live / parked) and stay on it (while searching, the first `Space` separates search terms and a second one toggles) |
 | `Tab` | toggle the highlighted row and advance to the next |
 | `Enter` | apply every staged change (with `--dry-run`: show the plan) |
 | `Esc` / `Ctrl-C` | cancel; nothing is applied |
@@ -300,7 +337,8 @@ off under `--dry-run`. Esc closes the prompt.
 Nothing happens while the picker is open. Changes are staged, the screen is
 torn down, and only then do the real operations run, so their output (which
 companion files moved, which were kept because they are shared) is readable
-instead of fighting curses for the terminal.
+instead of fighting curses for the terminal. Claude Code sessions running the
+[pane mod](#claude-code-pane-mod) pick up the Claude changes within 5 s and reload them.
 
 Built on stdlib `curses`, so there is nothing to install on macOS and Linux
 (on Windows, `pip install "agent-toggle[windows]"` pulls `windows-curses`).
@@ -352,10 +390,10 @@ end of input, `q` or an empty line exits `0`
 
 | group | rows |
 |---|---|
-| General | Check for updates, Color, Language, Default harness |
+| General | Check for updates, Color, Logo, Language, Default harness |
 | Harnesses | one On/Off per harness, with `found ~/.x` or `not on this machine` |
 | Picker | default sort, harness filter, type filter |
-| Notifications | Telegram bot token (masked), Telegram chat ID |
+| Notifications | Telegram bot token (masked, last 4 characters shown), Telegram chat ID |
 | Tools | Health check (`doctor`; then asks y/n per fix it can run), Install shims, Undo last change (asks y/n), Send test message, Sync CI secrets (asks y/n); their output shows in an overlay |
 
 Keys: `↑` `↓` move (wraps, skips headings), `←` `→` cycle a value, `Enter` / `Space` change or run
@@ -371,6 +409,7 @@ tool does not know are kept):
 |---|---|---|---|
 | `update_check` | `true` / `false` | `true` | `AGENT_TOGGLE_UPDATE_CHECK` |
 | `color` | `auto` / `always` / `never` | `auto` | `AGENT_TOGGLE_COLOR` |
+| `logo` | `color` / `mono` / `animated` / `off` | `animated` | `AGENT_TOGGLE_LOGO` |
 | `language` | `en` / `zh-TW` | `en` | `AGENT_TOGGLE_LANG` |
 | `default_harness` | a harness name | `claude` | `AGENT_TOGGLE_DEFAULT_HARNESS` |
 | `harness.<name>` | `true` / `false`, for `claude` `codex` `grok` `opencode` `openclaw` `copilot` `vibe` `devin` `agy` | `true` | |
@@ -427,6 +466,13 @@ options and examples; `help help` is the overview, `help version` shows how to p
 and an unknown command exits `2`. It honours `--color` and `NO_COLOR`, and is
 translated when `language` is `zh-TW`.
 
+The overview opens with the agent-toggle logo (cyan to blue to magenta): the large one from 97
+columns, a compact one from 49, none below that. Piped output gets the plain glyphs, a stream
+that cannot encode them gets none. The `logo` setting picks `color`, `mono` (one colour),
+`animated` (one light sweep, only on a colour terminal) or `off`. The config menu shows the same
+logo, centred, when the window is tall enough to fit the whole menu beside it; there,
+`animated` sweeps on entry and again after every 5 s without a key, and any key stops it.
+
 ## What each harness supports
 
 | harness | home | skill | agent | command | rule | plugin | mcp |
@@ -523,7 +569,8 @@ visible rather than silent.
   hand-rolled serializer would mangle comments — a text slice is lossless and
   much less code.
 
-MCP changes need a **new session** to take effect.
+MCP changes need a **new session** to take effect, except in a Claude Code session running the
+agent-toggle mod, which runs `/mcp disable|enable <name>` for you (see the mod section).
 
 ### Claude MCP scopes
 
@@ -554,6 +601,8 @@ agent-toggle profile diff work                       # what apply would do; writ
 agent-toggle profile apply ~/dotfiles/agent-toggle/work.json --dry-run
 agent-toggle profile apply work
 agent-toggle profile list
+agent-toggle profile rename work flutter             # -> ~/.agent-toggle/profiles/flutter.json
+agent-toggle profile delete flutter
 ```
 
 **`apply` toggles only the items the profile mentions.** An item the profile
@@ -582,6 +631,14 @@ one batch, logged, undoable with `undo`).
   so a project profile can never disable the user's items.
 - `save` skips (with a warning) any item whose name `apply` would refuse, such as
   an MCP server called `team/search`, so a saved profile always applies.
+- `save` to an existing name overwrites it with the live state now; that is how a
+  profile is updated.
+- `delete <name>` and `rename <old> <new>` work on stored profiles only (a file
+  you saved with `--out` is yours to move or remove) and change no item. Both exit
+  `2` for a name with no stored profile or a path; `rename` also exits `2` when
+  `<new>` already exists (it never overwrites) or is not a valid profile name, and
+  keeps the file byte for byte. Neither takes `--dry-run` or `--project`: a
+  project profile is renamed or deleted by its name like any other.
 
 ## Undo and `enable --all`
 
@@ -761,6 +818,7 @@ bookkeeping.
 | `state.json` | current disabled list (schema v3; atomic write, mode `0600`), saved before each flag write or dir move with the op in flight under an optional `pending` key, and after each MCP / plugin item |
 | `lock` | held by `disable` / `enable` / `enable --all` / `undo` / `profile apply` / `migrate` / `ui` for the whole batch; a second run waits 5 s then exits `3` (stale after 10 min *and* its PID is gone; a live batch refreshes it per item; the error says when its PID is gone, i.e. a killed run left it) |
 | `log.jsonl` | one line per operation (mode `0600`), see below |
+| `reload.json` | the Claude rows the last `ui` Enter changed, rewritten each time; open Claude Code sessions running the mod poll it and reload them |
 | `mcp-backups/` | `<harness>__<server>.json`, or `<sha8>__<harness>__<server>.json` for a project `.mcp.json` (mode `0600` -- may hold auth headers) |
 | `companions/` | parked exclusive helper files |
 | `parked/user/<owner>/<dir>/` | user-scope items, e.g. `parked/user/claude/skills/<name>` (`<owner>` = the harness whose home holds the dir; a dir shared by several harnesses parks once, under its owner); an absolute OpenCode `skills.paths` dir parks under `parked/user/opencode/ext-<sha8>/` (`<sha8>` of its resolved path). Nothing is ever parked inside a harness home |
@@ -838,6 +896,69 @@ deletion lines in `git status`.
 unresolved paths makes every companion look like it lives outside the harness.
 Both sides are resolved first.
 
+## Claude Code pane mod
+
+A Claude Code mod in `mod/` that shows the picker as a pane inside Claude (not to be confused
+with the [Mods](#mods) group the picker lists). It is driven by the installed `agent-toggle`
+CLI (`cost --json`, `disable|enable --json`) and works for the claude harness only; the curses
+`agent-toggle ui` stays the cross-harness UI.
+
+Install it with `/plugin install agent-toggle --marketplace weskao/agent-toggle` (see
+[Install](#install)); `claude plugin update agent-toggle@agent-toggle` picks up a new release. To
+run your own checkout instead (edits reload as you save):
+
+```sh
+claude --plugin-dir ./mod    # then run /agent-toggle inside Claude
+```
+
+- `/agent-toggle` or `/agent-toggle open` opens the pane, `/agent-toggle close` closes it,
+  `/agent-toggle toggle` does whichever applies; any other argument that is not a CLI command (below) prints the usage. Ctrl-X X
+  also closes it. Claude can do the same when you ask it to open or close agent-toggle: the mod
+  registers a `pane` tool (`mcp__agent-toggle__pane`, `action`: `open` | `close` | `toggle`),
+  e.g. "open agent-toggle", "close the agent-toggle pane", "toggle the agent-toggle pane".
+  Turning resources on or off is unchanged: toggle a row in the pane, or ask Claude in plain
+  words ("disable the demo-skill skill", "add it back"), which runs the CLI through the skill shim.
+- It looks like the `ui` picker: a title with the live-token total, a search line, rows under
+  type headings with counts, green `●` live / yellow `○` parked, tokens and a bar per row (a
+  parked row shows `(N)`, what restoring it would load), the cursor row in reverse video, and a
+  key-chip footer. The list scrolls with the cursor; the status line shows its position (`7/23`),
+  and the first and last list lines count the rows out of view (`↑ 4`, `↓ 12`).
+- The mouse wheel over the list moves the cursor, as in the terminal picker; no click needed.
+- Click the list once so it takes keys (Esc hands the keyboard back to the prompt). Then `↑` `↓`
+  `PgUp` `PgDn` `Home` `End` (also `Ctrl-P` / `Ctrl-N`) move; `Space` or `Enter` toggles the row,
+  `Tab` toggles and moves on; `/` or any other letter starts a search (terms ANDed over
+  `type/name` and the group, letters-in-order if nothing matches; a second `Space` toggles);
+  `s` sorts by name or cost; `n` toggles every type with this row's name (skill, command, MCP,
+  ...), like `disable all <name>`; `Ctrl-U` clears the search. A click moves the cursor.
+- A toggle applies at once (no staging, no Enter to apply); the toast names the target and the state
+  with an emoji and a bracketed tag (no colour codes, so nothing garbles; on the Linux console and the
+  legacy Windows console, where emoji draw as boxes, the tag alone): `✅ [ENABLED] skill: foo`, `⛔ [DISABLED] mod: bar (/reload-plugins)`,
+  `⛔ [DISABLED] rule: baz (takes effect in a new session)`, `⛔ [DISABLED] all: foo` after `n`, or `❌ [FAILED] disable skill: foo - <error>`.
+  The row flips the moment you press (and flips back
+  if the CLI fails); the list reloads in the background. While a toggle runs, further keys are ignored.
+- Opening the pane shows the same braille spinner and `Loading resources…` as `agent-toggle ui`
+  until the list arrives, and a spinner beside `working…` while a toggle runs.
+- Not in the pane: harness tabs (claude only), and the `t` `p` `a` keys. Where Claude draws no
+  such pane (the mobile app, VS Code) it is a plain list of buttons instead.
+- `/agent-toggle <command> ...` with `disable`, `enable`,
+  `undo`, `list`, `status`, `cost` or `doctor` runs the CLI instead of opening the pane and prints
+  one line per result row: `/agent-toggle disable skill foo bar` (it adds `--json`, and
+  `--harness claude` unless you pass `--harness`). A toggle reloads an open pane.
+- Inside Claude Code the mod then applies the change to the running session by queueing the slash
+  command for you, once per kind: skills and commands `/reload-skills`, plugins (mods included)
+  `/reload-plugins`, an MCP server `/mcp disable|enable <name>`. The toast names it, e.g.
+  `⛔ [DISABLED] skill: foo (/reload-skills)`. Agents and rules have no reload and still say
+  `takes effect in a new session`. If Claude refuses a queued command, a toast says so.
+- `agent-toggle ui` in another terminal applies on Enter and writes the Claude rows it changed to
+  `~/.agent-toggle/reload.json`. Every open session running the mod checks that file's mtime every
+  5 s (one `stat`) and queues the same reloads, toasting `agent-toggle ui: /reload-skills`. A file
+  older than the session is ignored.
+- Needs `agent-toggle` on `PATH`.
+- On Windows `agent-toggle` must resolve to a real executable on `PATH` (a pip or uv install
+  provides `agent-toggle.exe`).
+
+The design is in `docs/MOD.md`.
+
 ## Design and roadmap
 
 Architecture, harness survey, cost model, known gaps and the phased roadmap
@@ -855,6 +976,13 @@ ruff check .
 stdlib `unittest`, no fixtures, no network. Every test runs against a
 throwaway temp `HOME` and a stubbed `claude` CLI.
 
+The pane mod has its own checks (they need the `claude` CLI, so CI does not run them):
+
+```sh
+claude plugin validate mod
+claude plugin test mod
+```
+
 `python3 tools/wheel_smoke.py` builds the wheel, installs it without extras into a
 fresh venv (what a PyPI user gets) and runs the CLI against a temp `HOME`; CI runs
 it on macOS, Linux and Windows.
@@ -897,6 +1025,8 @@ prints a `::notice::` and exits 0, so it never adds a second red X.
 ## Releasing
 
 A release is a git tag `vX.Y.Z` (where `X.Y.Z` must match `agent_toggle.__version__`).
+Bump `version` in `mod/.claude-plugin/plugin.json` to the same value by hand; nothing checks it,
+and without it `claude plugin update` does not hand marketplace installs the new mod.
 Pushing the tag triggers `.github/workflows/release.yml`:
 
 1. **Build** job checks that the tag matches `__version__`, builds a wheel and

{agent_toggle-0.8.1 → agent_toggle-0.10.0}/agent_toggle/__init__.py

@@ -1,3 +1,3 @@
 """agent-toggle: temporarily disable / re-enable AI-agent resources."""
 
-__version__ = "0.8.1"
+__version__ = "0.10.0"