git:20260930.2f50853 to git:20261012.38fa26b

- > Manager mode has **TWO independent layers**. Keep them straight:

- >

- > 1. **SOFT codewords (`++m` / `++a` / `++rr` / `++r`) — autonomous, hook-driven, ALWAYS fire.** A `UserPromptSubmit` hook (`hooks/manager-prompt.mjs`) watches every prompt; when it sees a codeword it injects the matching block as `additionalContext` for that one turn. This is NOT enabled/disabled by this skill — it works regardless of skill state. The skill only **explains** it (`status`) and **customizes its TEXT** (`edit`/`purge`).

- > Detection (longest-prefix first within the review group):

- > - `++m` → Manager: `permission_mode === 'plan'` selects `planmode` (full + read-only plan addon); otherwise `full`. Plan graph changes and cron creation for execution; write neither in Plan mode. No separate `++mp`.

- > - `++a` → Architecture-first directive (`architect`). Injects `[DIRECTIVE: ARCHITECTURE-FIRST]` before implementation — delegate an architecture pass that fits the project's existing architecture, patterns and rules; robust, scalable, and SIMPLE (no over-engineering); find the closest well-built counterpart in the repo and take its principles (additive to conventions/rules, not a replacement), clean seams. Independent group — combines with `++m` and the review group. Mode-agnostic: same block in plan and normal mode (in plan mode it is written into the plan).

- > - `++rr` → Regression Review discipline (`review-regression`) — after each significant phase: no regression + project standard + correctness; two-phase review→double-check→fix; final cross-review at task end. Tested before `++r`.

- > - `++r` → Review discipline (`review-double`) — two-phase multi-agent review→double-check→fix after each significant change; codeword-only (no ambient/wall injection).

- > - Bare `+++` → `cron-plan`, ONLY when `permission_mode === 'plan'`: plan a unique hourly session anti-drift cron per top-level task (user interval overrides), created during execution through the task board. Outside Plan mode it injects nothing; it does not arm/disarm the wall. Independent of, and combinable with, the other groups.

- > - When the HARD wall is ON, the Manager (full) block is ALSO auto-injected on EVERY turn — no codeword needed. Codewords and wall injection are independent.

- > 2. **HARD wall — opt-in, this skill only, project-only, persistent, default OFF; no codeword.** Install a self-contained PreToolUse guard into `<cwd>/.claude/settings.local.json`; it is not a plugin hook. Arming requires P1: explicit user wall-enable wording or "Yes, arm it now"; bare install/enable and autonomy never confirm. Declined install still installs disarmed; declined enable/one-shot changes nothing. The guard denies main-session mutations while subagents remain free (`agent_id`). Actions and exact policies are below.

- >

- > Either layer works independently: wall enforces delegation; codewords supply prompts.

- >

- > **INSTALL-ONCE + STATE-GATE:** register once in personal, gitignored `settings.local.json`; recheck project `state.json {hard}` at runtime. Armed walls deny arbitrary Edit/Bash, so `disable` must never edit settings: flip state through the self-exempt command below, valid even at strict. Registration remains inert while disarmed.

- >

- > **THE EXEMPT COMMAND (memorize this shape — nothing else gets through an armed wall):**

- > The guard exempts it only when ALL of these hold: the command starts with `node `, the FIRST argument after `node` **resolves (realpath) to the helper this project actually installed** — `<root>/.claude/brewtools/manager/manager-state.mjs`, or the plugin's own `hooks/lib/manager-state.mjs` next to the guard — there is no shell operator outside quotes, no `$` expansion, no eval flag (`-e`/`--eval`/`-p`/`--print`/`--input-type`/`--require`/`--import`/`--loader`), and the remaining arguments are the helper's own CLI (`get` | `set hard=<true|false> level=<strict|balanced> mcpAllow=<mcp__srv__tool[,...]|> [--cwd DIR]`). **No `BT_ROOT=` prelude, no `&& echo`, no `|| echo`, no `test -f` — every one of those is a shell operator and turns the exemption OFF.** A file merely *named* `manager-state.mjs` elsewhere on disk is NOT exempt: the anchor is the absolute installed path, not the filename or a path suffix. `install`/`upgrade` copy the helper into the project precisely so this command needs no path resolution.

- | Who is walled | The MAIN session only. Subagents are free BY DESIGN — the discriminator is `agent_id` in the PreToolUse payload, present only for subagent calls. A `claude --agent <name>` main session carries `agent_type` without `agent_id` and IS walled. |

- | Project root | Resolved as `CLAUDE_PROJECT_DIR` → upward walk for `.git`/`.claude` → hook `cwd`, plus the guard's own installed directory. State is found from ANY nested working directory; a deep `cwd` no longer silently disables the wall. |

- | Fail-closed | An unparseable PreToolUse payload, an internal guard error, or an installed manager directory whose `state.json` is missing/corrupt all DENY the main session (at `strict` semantics) instead of passing through. Subagents still pass. |

- | `balanced` Bash | A strict allowlist of exact binaries (`ls cat pwd which head tail wc grep rg date whoami basename dirname realpath test [ jq echo find git gh node`) plus per-binary flag vetting: `rg --pre/--pre-glob/--search-zip`, `find -exec/-ok/-delete/-fprint*`, `git -c/--exec-path/--upload-pack/--ext-diff`, and `node` anything other than `--check` are DENIED. `env` is not on the list at all — it is a universal exec wrapper. Any `>`/`<` redirection, `$(...)` or backtick anywhere denies the whole command. |

- | `strict` Bash | Everything above is denied too; only the exempt state CLI runs. |

- | MCP | Classified on the tool segment after the second `__`, so a server named `search` cannot launder `mcp__search__destroy_all`. Default-deny per token: an unrecognised verb **or noun** → denied, and an ambiguous verb (`query`/`resolve`) reads ONLY inside a docs/reference name (`query-docs`, `resolve-library-id`), so `mcp__sqlite__query`, `mcp__sqlite__query_table` and `mcp__linear__resolve_issue` are all denied — those write on a DB server or tracker. `mcpAllow` is the escape hatch. |

- | `mcpAllow` | Optional state key, the escape hatch for a false MCP denial. Entries are an exact scoped name `mcp__server__tool` or a whole-server prefix `mcp__server__*`; consulted BEFORE the classifier and at **`balanced` ONLY** — `strict` denies all MCP, allowlisted or not. A malformed value allows nothing and never breaks state. |

- 4. Two residual cases need action OUTSIDE the session, and there is no in-session workaround — do not go hunting for one:

- - Claude Code changes the PreToolUse payload shape so the guard cannot parse it. Every main-session mutation is then denied. Fix: quit and delete the `brewtools-manager-guard` entry from `.claude/settings.local.json` in an editor.

- - Deleting the whole `.claude/brewtools/` tree disarms the wall (no manager directory = never installed). That is the documented consequence of a manual `rm -rf`, not a way to disable the wall — use `disable` or `uninstall`, which keep settings and files consistent.

- Position 1 of `$ARGUMENTS` is a **free-form prompt** (RU/EN) — modes and flags are optional and may

- follow in any order. Nobody types keys: resolve the action + scope FROM the prompt via

- 1. Strip flags. An explicit action token anywhere wins outright, no scoring.

- 2. Else score actions by distinct whole-word keyword hits (P0 table). Highest unique score wins.

- Tie with a destructive action (`purge`) -> `AskUserQuestion`; tie with `status` -> `status`;

- tie of two mutating actions -> the keyword appearing first; no hits -> `inline-run` for a real task, otherwise `status`.

- 3. Empty arguments -> `status`; ask ONE scoping `AskUserQuestion` only when the answer changes

- what gets written or armed. `status` asks nothing.

- 4. Outcome-changing ambiguity (incl. `hard-one-shot` vs `manager-run`, enable vs disable) -> ONE

- `AskUserQuestion` (max 4 questions) BEFORE any work — this is P1.

- 5. Prose that is not an action/id/path is still input: extract the task from `<task> в хард

- Then print this block ONCE, after P1 and before the first action (P2). A read-only `status`

- prints it immediately before its report:

- Labels are literal; authored plan values are English (preserve INPUT verbatim). `install`/`upgrade`/`enable`/

- `disable`/`uninstall`/`purge`/`level` touch the HARD-wall layer; `edit` touches the codewords

- | Every Bash call ends with `&& echo "✅ ..." \|\| echo "❌ FAILED ..."` | ALL **except** the bare exempt state-write command (see the crux box) — appending `&& echo` there makes the armed wall deny it |

- | The HARD wall (`state.hard`) is **PROJECT scope ONLY** — there is no global wall. Always `writeState('project', ...)` for `hard`/`level` | install/enable/disable/level/hard-one-shot |

- | The wall is **installed INTO the project**, not shipped as a plugin hook. `install`/`upgrade` copy the guard + register it in `<cwd>/.claude/settings.local.json`; `enable`/`disable` flip state only; `uninstall`/`purge` deregister | install/upgrade/enable/disable/uninstall/purge |

- | All `settings.local.json` mutations go through a **node Bash block** (read-merge-atomic-write), NEVER the Edit tool — the Edit tool may be blocked by an armed wall, and we must not depend on it | install/upgrade/uninstall/purge |

- | State writes go through `writeState(scope, partial, cwd)` (atomic: lockfile + tmp + rename) — never write `state.json` by hand | P2 |

- | State reads go through `resolveState(cwd)`; prompts via `resolvePrompt(mode, cwd, root)` / `resolvePromptPath(scope, mode, cwd)` | P2, status |

- | Never reimplement resolution logic — always call the helpers | ALL |

- | GLOBAL prompt-override paths (`~/.claude/manager/prompts/*`) are PROTECTED for Write/Edit — write ONLY via the Node helper through Bash. Project prompt overrides are plain writes (still prefer helper) | edit/purge |

- | Thing | Scope | Files |

- |-------|-------|-------|

- | **Wall state** `{hard, level}` + optional `mcpAllow` (runtime kill-switch) | **PROJECT ONLY** | `<cwd>/.claude/brewtools/manager/state.json` |

- | **Wall registration** (persistent plumbing) | **PROJECT ONLY** | `<cwd>/.claude/settings.local.json` (PreToolUse `*` entry) + copied guard `<cwd>/.claude/brewtools/manager/hardmode-guard.mjs` |

- | Soft default `mode` field (informational) | project state | same `state.json` |

- | **Prompt-text overrides** (`edit`/`purge`) | project **or** global (separate files) | project: `<cwd>/.claude/brewtools/manager/prompts/<mode>.md` · global: `~/.claude/manager/prompts/<mode>.md` |

- > "Wall scope" is fixed (project). "Prompt-text override scope" is a different, independent axis that `edit`/`purge` may target globally. Do not let `--scope global` leak onto the wall — it has no meaning there.

- The plugin root is resolved from the skill's OWN directory (the `CLAUDE_SKILL_DIR` prompt substitution), never from `CLAUDE_PLUGIN_ROOT` -- that env var is not exported to a skill's Bash tool. Every Bash block resolves `BT_ROOT` this way (no hardcoded version):

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- - State helper: `$BT_ROOT/hooks/lib/manager-state.mjs` — exports `resolveState`, `writeState`, `resolveStatePath`; also a CLI: `node <path>/manager-state.mjs get|set hard=<true|false> level=<strict|balanced> mcpAllow=<mcp__srv__tool[,...]|> [--cwd DIR]`

- - **Guard source (shipped, self-contained, NOT in plugin `hooks.json`):** `$BT_ROOT/hooks/hardmode-guard.mjs` — `install`/`upgrade` copy this into the project

- - Plugin default prompt inventory: `$BT_ROOT/skills/manager-setup/references/<mode>.md` for `full`, `planmode`, `cron-plan`, `architect`, `review-regression`, `review-double`. `install`/`upgrade` deploy only wall assets; prompt defaults remain plugin-owned and refresh with the plugin. Preserve project/global overrides; status/show/explain resolves all six.

- - Wall policy + canonical status text: `$BT_ROOT/skills/manager-setup/references/hard.md` — **Read it for the install model, status explainer and the allowlist details.**

-

- Project install targets (resolved from `process.cwd()`):

- - Copied guard: `<cwd>/.claude/brewtools/manager/hardmode-guard.mjs`

- - Copied state helper: `<cwd>/.claude/brewtools/manager/manager-state.mjs` — the off-switch CLI. Copied so `disable`/`level` are a fixed path needing no `BT_ROOT` resolution (resolution needs shell operators, which the armed wall denies)

- - Registration: `<cwd>/.claude/settings.local.json` — a `PreToolUse` matcher `"*"` entry whose command runs `node <ABS path to copied guard>`. Tagged with marker `brewtools-manager-guard` so `uninstall`/`purge` can find it.

-

- ### Resolution chains (must match helpers exactly)

-

- | What | project | → global | → default |

- |------|---------|----------|-----------|

- | State `mode` (informational) | `<cwd>/.claude/brewtools/manager/state.json` | `~/.claude/manager/state.json` | `mode:'full'` |

- | Wall flags `{hard, level, mcpAllow}` | `<cwd>/.claude/brewtools/manager/state.json` | (no global — PROJECT-ONLY) | `{hard:false, level:'balanced', mcpAllow:[]}` |

- | Prompt text `<mode>` | `<cwd>/.claude/brewtools/manager/prompts/<mode>.md` | `~/.claude/manager/prompts/<mode>.md` | `$BT_ROOT/skills/manager-setup/references/<mode>.md` |

-

- > The wall flags (`hard`/`level`) are resolved **PROJECT-ONLY in code** — the global `state.json` does NOT enable the wall. The skill writes them to **project** scope only. (The informational `mode` field may still resolve from global; `hard`/`level` do not.)

- Parse `$ARGUMENTS` (or the user's NL prompt, RU+EN) into `{ action, scope, mode, level, task }` using `references/intent-routing.md` — **Read and follow it**.

- Actions, canonical order: `status`, `install`, `upgrade`, `enable`, `disable`, `uninstall`, `purge`, plus the extras `level <strict|balanced>`, `edit`, and the run actions `hard-one-shot`, `manager-run`, `inline-run`.

- > `on` / `off` / `reset` / `setup` / `remove` are REMOVED as command words. `on` and `off` survive only as free-text synonyms routed to `enable` / `disable` above; `reset` routes to `purge`. Never print them as commands.

-

- > **Arming is never automatic.** `install`, `enable`, and `hard-one-shot` are the only actions that can

- > write `state.hard=true`. Resolving the action (even from an explicit keyword like `install`, or

- > autonomy phrasing like "autonomous, decide everything yourself"/"автономно выбирай сам") is NOT

- > confirmation to arm — that confirmation is a separate, mandatory P1 gate. See P1 below.

- Prompt-text override scope (ONLY for `edit`/`purge`): default = `project`. `--scope global` OR `глобально` / `globally` → `global`. This scope does NOT apply to `install`/`upgrade`/`enable`/`disable`/`uninstall`/`level` (those are project-only).

- If the action is ambiguous or signals conflict (e.g. enable + disable, a task that might be `hard-one-shot` vs `manager-run`, control implied but no verb) → `AskUserQuestion` with the candidate actions as options. Otherwise proceed.

-

- > `hard-one-shot` flips the wall and auto-reverts; `manager-run` leaves it untouched. Conflicting/unclear markers require a question; a task with no marker uses `inline-run`.

- **Arm-confirmation gate (unconditional — not only on ambiguity).** `install`, `enable`, and

- `hard-one-shot` are the only actions that can write `state.hard=true`. Before P2 runs any of them,

- check whether the user's OWN prompt already carries EXPLICIT confirming wording: the words "hard

- wall" / "хард уолл" / "стену" TOGETHER with an enable/arm verb ("enable"/"arm"/"включи"/"заarmи"),

- e.g. "enable the hard wall", "arm the wall", "включи Hard Wall", "включи хард уолл", "заarmи стену".

- The bare skill verb alone (`install`/`установи`/`enable`/`включи` with no "wall" wording) does

- **not** count, and no autonomy-permission phrasing ever counts ("autonomous, decide everything

- yourself", "автономно", "выбирай сам" — these NEVER satisfy the gate). If explicit wording is

- present, treat the arm as already confirmed and skip the question. Otherwise ask exactly ONE

- `AskUserQuestion` before P2:

- - `install`: "Arm the HARD wall in this project now? It will block Write/Edit/Bash in the MAIN

- session (subagents stay free) until disabled." — options `Yes, arm it now` / `No, just install

- (stay disarmed)`.

- - `install` proceeds either way — it always copies the guard and registers the hook; `ARM` only

- decides whether `state.hard` is written `true` or `false`. `ARM=false` → report "installed but

- NOT armed — run `enable` (or confirm next time) to arm it."

- - `enable` / `hard-one-shot`: `ARM=false` aborts the whole action before running anything — do not

- touch `state.hard`, report plainly that nothing changed.

- Print the `## Prompt contract` PLAN block first — INPUT/MODE from P0, SCOPE naming which layer

- (codewords vs HARD wall, or both for `hard-one-shot`) — before running the mapped section below.

- `status` prints the same block immediately before its report instead of before a mutation.

-

- > Sections below map 1:1 to `action`: `install`, `upgrade`, `enable`, `disable`, `uninstall`, `purge`, `level`, `status`, `edit`, and the three run actions.

- `install` is a five-step sequence: (1) set arm state per the P1 confirm gate (`ARM`), (2) copy the guard into the project, (3) idempotently register it in `settings.local.json`, (4) turn the task-graph tools on in that same file, (5) report whether a `/reload` is needed. All five run in ONE node Bash block so the registration is atomic and self-contained. **`ARM` (`true`/`false`) is resolved by the P1 arm-confirmation gate — substitute it into the block below before running, same as `LEVEL` in the `level` action.** The block:

- - sets `state.hard = (ARM === 'true')` via `writeState('project', {hard:arm})` — arms only when P1 confirmed,

- - copies `$BT_ROOT/hooks/hardmode-guard.mjs` → `<cwd>/.claude/brewtools/manager/hardmode-guard.mjs` **and** `$BT_ROOT/hooks/lib/manager-state.mjs` → `<cwd>/.claude/brewtools/manager/manager-state.mjs` (both overwritten on EVERY `install`, so plugin updates propagate; the second one is the off-switch CLI),

- - read-merge-atomic-writes `<cwd>/.claude/settings.local.json`, adding a `PreToolUse` matcher `"*"` entry that runs `node <ABS copied-guard path>` tagged `brewtools-manager-guard`, but ONLY if no entry already points at the manager guard (idempotent — running twice = ONE entry),

- - in that SAME merge sets `env.CLAUDE_CODE_ENABLE_TODO_TOOLS = "1"` when Claude Code is >= 2.1.233 — creating the `env` object if absent, preserving every other key. This is unconditional and never asks: from 2.1.233 `TaskCreate`/`TaskUpdate`/`TaskGet`/`TaskList` are gated OFF by default, and the manager framework has no task graph without them. Below 2.1.233 the var does nothing and the tools are on anyway, so the write is skipped and the block says so,

- - prints `newlyRegistered` and `todoTools` so you know whether to surface the `/reload` note and what happened to the task tools.

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { echo "❌ cannot resolve project root — looked for CLAUDE_PROJECT_DIR, git toplevel, then .git/.claude above $PWD; nothing written"; exit 1; }

- ARM=<true|false>

- import fs from 'node:fs'; import path from 'node:path';

- const todoTools = todoToolsGated ? 'enabled (CC ' + ccVer + ')'

- // 1. arm only if P1 confirmed

- await writeState('project', {hard:arm}, cwd);

- // 2. copy guard + off-switch CLI (overwrite each install)

- fs.mkdirSync(dir, {recursive:true});

- fs.copyFileSync(src, guard);

- fs.copyFileSync('${BT_ROOT}/hooks/lib/manager-state.mjs', helper);

- // 3. idempotent register, under the settings lock

- let newlyRegistered = false;

- arr.push({ matcher:'*', hooks:[{ type:'command', command:\`node \"\${guard}\" # \${TAG}\`, timeout:5 }] });

- if (todoToolsGated) { cfg.env = (cfg.env && typeof cfg.env==='object' && !Array.isArray(cfg.env)) ? cfg.env : {}; cfg.env.CLAUDE_CODE_ENABLE_TODO_TOOLS = '1'; }

- if (fs.existsSync(settings)) fs.copyFileSync(settings, settings + '.bak');

- const tmp = settings + '.tmp';

- fs.writeFileSync(tmp, JSON.stringify(cfg, null, 2) + '\n', 'utf8');

- fs.renameSync(tmp, settings);

- } finally { fs.rmSync(lock, {recursive:true, force:true}); }

- > **Root, lock, backup — the three invariants every settings-writing block here shares.**

- > `ROOT` is the canonical recipe (`CLAUDE_PROJECT_DIR` → `git rev-parse --show-toplevel` → upward

- > walk for `.git`/`.claude` → abort). An installer that cannot name its root ABORTS non-zero and

- > says what it looked for — it never writes to a guessed root, and it never uses raw `$PWD`, which

- > drifts to whatever subdirectory the session wandered into. The `settings.local.json.lock`

- > directory is an `O_EXCL` mutex (stale-broken after 30 s) so two setup skills running in parallel

- > cannot lose each other's edit; the file is re-read INSIDE the lock. `settings.local.json.bak` is

- > written before every rename. A read that is not `ENOENT`, or content that is not a JSON object,

- > ABORTS before anything is staged — a malformed settings file is never "the file is empty".

- - If `armed:true` → tell the user the exit command verbatim, with the real absolute path: `node <ABS project root>/.claude/brewtools/manager/manager-state.mjs set hard=false` — or just `/brewtools:manager-setup disable`, which runs exactly that.

- - If `armed:false` → tell the user plainly: **installed but NOT armed** — the guard is in place but `state.hard=false`; run `/brewtools:manager-setup enable` (or answer "Yes, arm it now" next time) to arm it.

- - If `newlyRegistered:true` → tell the user verbatim: `Hook installed in .claude/settings.local.json — run /reload (or restart the session) for the wall to take effect.`

- - If `newlyRegistered:false` → the entry already existed; a reload is only needed if `armed:true` just flipped a previously-disarmed state.

- - Report `todoTools` in one line: `enabled` → say `TaskCreate/TaskUpdate/TaskGet/TaskList enabled via env.CLAUDE_CODE_ENABLE_TODO_TOOLS in .claude/settings.local.json`; `skipped` → print the reason verbatim and move on.

-

- > The command in the registered entry uses an ABSOLUTE path to the copied guard and a `# brewtools-manager-guard` tag comment so `uninstall` can find it. Scope is always `project` — there is no global wall, never pass `'global'`.

- `upgrade` asks nothing: re-copy current `hardmode-guard.mjs` and `manager-state.mjs`, restore a missing registration, and merge `env.CLAUDE_CODE_ENABLE_TODO_TOOLS="1"` on Claude Code >= 2.1.233. This backfills older installs missing the off-switch CLI or task-tool key; run it after updating brewtools. The six default prompts above refresh with the plugin; overrides remain unchanged.

-

- > Restamp only `state.json` metadata through `writeState('project', {}, cwd)` (empty partial; docsync-setup upgrade pattern). Merge `{...existing,...partial}` preserves `hard`, `level`, `mode`, and unknown keys while refreshing `version/generated_by/last_updated`. Report `armStatePreserved` + `stateRestamped`, replacing historical `stateUntouched`. Setup-status row 8 reads top-level state version first, guard `brewcode-meta:` only if absent; not restamping leaves a permanent stale verdict. See `brewdoc/skills/docsync-setup/SKILL.md` upgrade.

- It ABORTS when the project has no wall installed. `upgrade` must never be a back door that arms a wall the user never asked for — an uninstalled project is told to run `install`.

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { echo "❌ cannot resolve project root — looked for CLAUDE_PROJECT_DIR, git toplevel, then .git/.claude above $PWD; nothing written"; exit 1; }

- import fs from 'node:fs'; import path from 'node:path';

- const todoTools = todoToolsGated ? 'enabled (CC ' + ccVer + ')'

- let newlyRegistered = false;

- if (!arr.some(has)) { arr.push({ matcher:'*', hooks:[{ type:'command', command:\`node \"\${guard}\" # \${TAG}\`, timeout:5 }] }); newlyRegistered = true; }

- if (todoToolsGated) { cfg.env = (cfg.env && typeof cfg.env==='object' && !Array.isArray(cfg.env)) ? cfg.env : {}; cfg.env.CLAUDE_CODE_ENABLE_TODO_TOOLS = '1'; }

- if (fs.existsSync(settings)) fs.copyFileSync(settings, settings + '.bak');

- } finally { fs.rmSync(lock, {recursive:true, force:true}); }

- // Restamp the metadata trio ONLY — empty partial, so hard/level/mode and every

- // unknown key merge through from the existing file untouched.

- const w = await writeState('project', {}, cwd);

- Surface the `/reload` note only when `newlyRegistered:true`. Report `todoTools` in one line the same way `install` does — an old project that predates the key gets it backfilled here.

- **Gated by P1.** This block only runs if the P1 arm-confirmation gate resolved `ARM=true` (explicit

- wording in the user's prompt, or an `AskUserQuestion` answered "Yes, arm it now"). If P1 resolved

- `ARM=false`, do NOT run this block at all — report plainly that arming was declined and nothing

- changed, and stop.

-

- `enable` flips `state.hard=true` and nothing else. If the wall was never installed there is no guard to arm, so the block reports `notInstalled` instead of writing a state that no hook reads — route the user to `install`.

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { echo "❌ cannot resolve project root — looked for CLAUDE_PROJECT_DIR, git toplevel, then .git/.claude above $PWD; nothing written"; exit 1; }

- `notInstalled:true` → tell the user the wall is not installed in this project and run `install` (which installs AND arms). Do not silently pretend it is armed.

-

- > This block is only reachable while the wall is OFF (that is the point of `enable`). If the wall is already armed, `enable` is a no-op — say so and skip the block rather than running a command the guard will deny.

-

- `disable` flips `state.hard=false` and does NOTHING else. It must NOT edit `settings.local.json`: while the wall is armed the guard DENIES `Edit`/`Bash` on arbitrary files. The ONE thing that gets through is the bare off-switch CLI — see the crux box at the top. So state is the runtime kill-switch; the registration stays harmlessly registered (the guard no-ops when `state.hard !== true`).

- It prints `{"file":...,"action":"written","state":{"hard":false,...}}` on success. That JSON on stdout is the ✅; a non-zero exit with a `manager-state:` line on stderr is the ❌.

-

- > **If it fails with `Cannot find module`** the project was installed by a brewtools older than the off-switch CLI. Do not try to resolve `$BT_ROOT` in the main session — that needs shell operators the armed wall denies. Instead delegate one subagent (`Agent`, always allowed, subagents bypass the wall) with: *"run `/brewtools:manager-setup upgrade` in this project, then `node <ABS_CWD>/.claude/brewtools/manager/manager-state.mjs set hard=false`"*.

- > The registered guard stays in `settings.local.json` and continues to fire on every tool call, but reads `state.hard` and immediately no-ops while disarmed. To remove the registration entirely, use `uninstall`.

- Removes the manager guard entry from `<cwd>/.claude/settings.local.json` (and the copied guard file). `state.json`, the prompt-text overrides and the `env.CLAUDE_CODE_ENABLE_TODO_TOOLS` key are **KEPT** — a later `install` comes back to the same `level` and the same customized prompt.

-

- > **Asymmetry, on purpose:** `uninstall` only deregisters the wall, so it must NOT touch `env.CLAUDE_CODE_ENABLE_TODO_TOOLS`. The task graph is useful with or without a wall, and silently switching `TaskCreate`/`TaskUpdate`/`TaskGet`/`TaskList` off while removing a hook would be a surprise. Only `purge` removes that key.

- **TWO separate Bash calls:** disarm with the bare exempt CLI, then deregister. Never merge them: the armed guard denies step 2's `BT_ROOT=`/`&& echo`/`node --input-type=module -e` shape; disarming lets it no-op.

- > Fail-closed ordering: **deregister -> re-read/confirm absence -> delete guard + helper**. Missing/corrupt installed state denies the main session; deleting files first leaves a broken registered hook. Delete failure restores the settings entry and exits nonzero. Repeated uninstall is idempotent: `deregistered:false`, exit 0.

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { echo "❌ cannot resolve project root — looked for CLAUDE_PROJECT_DIR, git toplevel, then .git/.claude above $PWD; nothing removed"; exit 1; }

- if (deregistered) {

- fs.copyFileSync(settings, settings + '.bak');

- } finally { fs.rmSync(lock, {recursive:true, force:true}); }

- After the block, tell the user: `Hook removed from .claude/settings.local.json — run /reload (or restart the session) to stop the guard from firing.`

- `purge` is `uninstall` plus every file this skill ever wrote: the whole `<cwd>/.claude/brewtools/manager/` tree (`state.json`, the copied guard, `prompts/`), the `env.CLAUDE_CODE_ENABLE_TODO_TOOLS` key that `install`/`upgrade` merged into `settings.local.json` (and the `env` object itself if that empties it), and — when the prompt-text scope is `global` — the global prompt override too. After a purge the project is indistinguishable from one that never ran this skill: `level` is back to `balanced` and the Manager prompt text is back to the plugin default.

- > **The task-graph key is removed HERE and only here.** `uninstall` deliberately leaves it: it deregisters the wall, and a project without a wall still wants `TaskCreate`/`TaskUpdate`/`TaskGet`/`TaskList`. `purge` is the "верни дефолт" verb, so it takes the key back out too. If the user only wanted the wall gone, they wanted `uninstall`.

- This is the only destructive action. Say what will be deleted BEFORE running it, and if the user only wanted the prompt text back on default, tell them so — there is no narrower verb any more.

- **EXECUTE** using Bash tool as a THIRD call, after both `uninstall` steps (step 1 disarms — without it this block is denied by the armed wall; step 2 deregisters). Substitute `SCOPE`:

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { echo "❌ cannot resolve project root — looked for CLAUDE_PROJECT_DIR, git toplevel, then .git/.claude above $PWD; nothing deleted"; exit 1; }

- import {resolvePromptPath} from '${BT_ROOT}/hooks/lib/manager-prompts.mjs';

- const removedDir = fs.existsSync(dir);

- if (removedDir) fs.rmSync(dir, {recursive:true, force:true});

- // Take the task-graph key back out — same read-merge-atomic-write, same lock as install.

- // uninstall keeps it on purpose; only purge restores the default.

- let todoKeyRemoved = false;

- if (!held) { console.error('ABORT: ' + lock + ' is held by another setup skill — the manager tree is gone but ' + settings + ' still carries CLAUDE_CODE_ENABLE_TODO_TOOLS; rerun purge'); process.exit(1); }

- if (cfg && typeof cfg === 'object' && !Array.isArray(cfg) && cfg.env && typeof cfg.env === 'object' && 'CLAUDE_CODE_ENABLE_TODO_TOOLS' in cfg.env) {

- delete cfg.env.CLAUDE_CODE_ENABLE_TODO_TOOLS;

- if (Object.keys(cfg.env).length === 0) delete cfg.env;

- fs.copyFileSync(settings, settings + '.bak');

- todoKeyRemoved = true;

- } finally { fs.rmSync(lock, {recursive:true, force:true}); }

- let globalPrompt = null;

- const g = resolvePromptPath('global', 'full', cwd);

- if (fs.existsSync(g)) { fs.unlinkSync(g); globalPrompt = g; }

- console.log(JSON.stringify({removedDir, dir, globalPrompt, todoKeyRemoved}));

- Report exactly what was deleted (or that nothing existed), including `todoKeyRemoved` — say plainly that the task-graph tools go back to the Claude Code default (OFF on 2.1.233+) and that `install` brings them back. Then run `status` — it will show `registered=no armed=OFF level=balanced` with `state source: default`.

- Same exempt CLI as `disable`, with `level=strict` or `level=balanced`. Controls the Bash/WebSearch/MCP policy of the wall while it is armed. `balanced` is the default. This is state-only — it never touches `settings.local.json`. See `references/hard.md` for the exact per-level allowlist.

- Status/show/explain is read-only: read merged state, resolve the six prompt modes above (including `cron-plan`), detect wall registration, and render the `references/hard.md` explainer with the additions below. It must teach:

- 1. **How `++m` works** — ALWAYS, per-turn, hook-driven (`manager-prompt.mjs`), independent of this skill. `++m` is plan-aware: it injects the planmode block (full + plan addon) when `permission_mode === 'plan'`, else the plain full block — there is NO separate `++mp` codeword. Show BOTH resolved blocks (full + planmode) so the user sees each variant. Also state: when the HARD wall is armed, the Manager (full) block is ALSO ambient-injected every turn with no codeword needed (codewords and wall injection are independent). The session-start banner is the other read-only plugin layer.

- 2. **The wall delivery model** — it is INSTALLED INTO this project, not a plugin hook: registered (once) in `<cwd>/.claude/settings.local.json` (personal, gitignored), gated at runtime by project `state.json {hard}`. Report BOTH: is it registered? is it armed (`hard`)?

- 3. **Current WALL state for THIS project** — `hard` armed/disarmed, `level` strict/balanced, and a brief allowlist summary (what main session may/may not do).

- 4. **How the verbs work** — `install` = install+arm (`/reload` only on FIRST install), `upgrade` = re-emit the guard with the arm state preserved and `state.json`'s metadata trio restamped to this plugin version, `enable` = arm an installed wall, `disable` = disarm only (registration kept), `uninstall` = deregister (state + prompt overrides kept, task-graph key kept), `purge` = uninstall + delete state, overrides and the task-graph key, `level` = strictness.

- 5. **Task-graph tools** — report whether `TaskCreate`/`TaskUpdate`/`TaskGet`/`TaskList` are actually available: the running Claude Code version, and whether `env.CLAUDE_CODE_ENABLE_TODO_TOOLS` is set in `<ROOT>/.claude/settings.local.json`, `<ROOT>/.claude/settings.json` or `~/.claude/settings.json`. Any layer counts — name the one that wins. Set nowhere on CC >= 2.1.233 → the tools are OFF and the manager framework has no task graph; give the one-line remedy.

- 6. **Anti-drift cron planning** — bare `+++` selects `cron-plan` only in Plan mode; planning creates no schedule or board state. Show its resolved source/text. Execution announces and creates a unique task-board session cron per top-level task, hourly unless overridden; task-specific methodology owns its tick instructions/ID/timezone/count/stop rule. Missing scheduler -> explicit gap; completion/cancellation -> stop and verify removal. `+++` does not control the wall.

- > **WHILE THE WALL IS ARMED, DO NOT RUN THE BASH BLOCK BELOW** — its `BT_ROOT=` prelude and `&& echo` tail are exactly what the guard denies. Build the same report with always-allowed tools instead:

- > - prompt blocks → `Read` the first path that exists, in order: `<cwd>/.claude/brewtools/manager/prompts/<mode>.md` → `~/.claude/manager/prompts/<mode>.md` → `$BT_ROOT/skills/manager-setup/references/<mode>.md`; that order IS the prompt source

- > - task-graph key → `Read` `<ABS_CWD>/.claude/settings.local.json`, then `<ABS_CWD>/.claude/settings.json`, then `~/.claude/settings.json` and look for `env.CLAUDE_CODE_ENABLE_TODO_TOOLS`. The CC version is in your session banner; if you cannot get it, report it as unknown rather than guessing

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- ROOT=$(if [ -n "$CLAUDE_PROJECT_DIR" ] && [ -d "$CLAUDE_PROJECT_DIR" ]; then printf %s "$CLAUDE_PROJECT_DIR"; elif r=$(git rev-parse --show-toplevel 2>/dev/null) && [ -n "$r" ]; then printf %s "$r"; else d=$PWD; while [ "$d" != "/" ]; do if [ -d "$d/.git" ] || [ -d "$d/.claude" ]; then printf %s "$d"; break; fi; d=$(dirname "$d"); done; fi)

- [ -n "$ROOT" ] || { ROOT=$PWD; echo "WARN: no project-root marker found; reporting on $PWD" >&2; }

- // `registered` is tri-state: true / false / null. null = the settings file exists but could

- // not be parsed, so registration is UNKNOWN — never report unknown as "not registered".

- // The key is legitimate in any settings layer, so probe all three and name the one that wins.

- let todoToolsLayer = null;

- ['~/.claude/settings.json', path.join(os.homedir(),'.claude','settings.json')]]) {

- try { const c = JSON.parse(fs.readFileSync(p,'utf8')); if (c && c.env && c.env.CLAUDE_CODE_ENABLE_TODO_TOOLS) { todoToolsLayer = label; break; } } catch {}

- ccVersion: ccVer || null, todoToolsGated, todoToolsLayer,

- > `stateVersion` is `null` on any state file written before the metadata keys existed, and on a project with no state file. `null` means UNKNOWN — never report it as up to date. `stale: true` -> recommend `upgrade`.

- >

- > `brewtools/hooks/lib/manager-state.mjs` writes `version` (plugin X.Y.Z), `generated_by:"brewtools:manager-setup"`, `last_updated` (YYYY-MM-DD), never `doc_type`, through writeState; resolveState preserves unknown keys. DEFAULT_STATE deliberately has no version, so absent/older raw state remains unknown instead of claiming current provenance.

- The four letter codewords fire on matching prompts regardless of skill state; `+++` additionally requires Plan mode. This skill never enables/disables their hooks.

- When armed, the main session cannot Write/Edit/WebFetch; it can delegate (Agent), read (Read/Grep/Glob), track (TodoWrite) and manage session timers (CronCreate/CronList/CronDelete). Task-board owns its timer matching/cleanup; scheduled work still passes through the wall. This exempts no external/OS scheduling. For Bash: strict denies all Bash; balanced permits read-only inspection.

- ## Task graph (TaskCreate/TaskUpdate/TaskGet/TaskList) — <ON|OFF>

- Claude Code <ccVersion>. From 2.1.233 these tools are gated off by default and need env CLAUDE_CODE_ENABLE_TODO_TOOLS=1.

- <when todoToolsLayer != null> ON — key set in <todoToolsLayer>.

- <when todoToolsLayer == null and todoToolsGated> OFF — the manager framework cannot build a task graph. Fix: /brewtools:manager-setup install (or upgrade if the wall is already installed).

- <when todoToolsLayer == null and not todoToolsGated> ON — this Claude Code predates the gate, the tools are available without the key.

- Install: /brewtools:manager-setup install (install+arm; /reload only on FIRST install)

- Uninstall: /brewtools:manager-setup uninstall (deregister from settings.local.json, then /reload)

- Operates on the Manager prompt text (internal mode `full`). If no project/global override exists for that scope, copy the current effective text into the override path, then print the path + content for the user to edit. This changes only what the codewords inject — it never touches the wall.

- **EXECUTE** using Bash tool (substitute `SCOPE`):

- SD="${CLAUDE_SKILL_DIR}"

- if [ -n "$SD" ] && [ -f "$SD/../../.claude-plugin/plugin.json" ]; then BT_ROOT=$(cd "$SD/../.." && pwd); else BT_ROOT=$(ls -d ~/.claude/plugins/cache/claude-brewcode/brewtools/*/ 2>/dev/null | sort -V | tail -1 | sed 's:/*$::'); fi

- [ -n "$BT_ROOT" ] || { echo "ERROR: cannot locate brewtools plugin root -- install/update brewtools first."; exit 1; }

- const cwd = process.cwd(); const root = '${BT_ROOT}';

- > `SCOPE` here is the prompt-text override scope (`project` default, or `global`), NOT the wall. Global override goes under `~/.claude/manager/prompts/` (protected) — must go through this Node block. Tell the user the path; they (or you, for project scope) edit it with the Edit tool.

- The user gave a REAL task plus a hard-mode marker. **Gated by P1**: this whole action only proceeds

- if the P1 arm-confirmation gate resolved `ARM=true` (explicit wording, or "Yes, arm it now"). If P1

- resolved `ARM=false`, abort the arm entirely — do not turn the wall on, do not run the task under

- it, report plainly that nothing changed (suggest `manager-run` if they want the task done without

- the wall). Otherwise, run it once under the wall, then auto-revert:

-

- 1. **Turn the wall ON** — run the `install` block with `ARM=true` (it installs the guard + the off-switch CLI and arms), or the `enable` block if the wall is already installed.

- 2. **Act as Manager** for the task: resolve the `full` block, treat it as your operating contract, **build a TaskGraph** (`TaskCreate`/`TaskUpdate`) and **delegate** to the best-matching expert agent(s) in parallel where independent. **Never implement by hand** — and with the wall ON, mutating tools are denied anyway.

- 3. **Auto-revert** — when the task is complete, **turn the wall back OFF** with the `disable` command, VERBATIM (the wall is armed at this point, so nothing else gets through): `node <ABS_CWD>/.claude/brewtools/manager/manager-state.mjs set hard=false`.

- > **Revert on failure too.** If the task aborts, errors, or you stop early, STILL run that exact command so the wall does not silently persist beyond the one-shot. The wall must end OFF exactly as it started, regardless of outcome. State the revert explicitly to the user.

- Run the task in Manager role WITHOUT touching the wall — discipline by prompt only:

-

- 1. Resolve the `full` block text (`resolvePrompt('full', ...)`).

- 2. PREPEND it as your operating contract.

- 3. **Build a TaskGraph** and **delegate** to the best-matching expert(s), fan out independent work in parallel. **Never implement by hand.**

- > Wall state is left exactly as it was (could be on or off). No `writeState` here.

- Same as `manager-run`: prepend the `full` block, build a TaskGraph, delegate, never implement by hand, do not touch the wall. This is the gentle default for a bare task.

-

- > When the wall is ON the full block is already ambient-injected by the hook; the skill still prepends it for consistency (one-shot runs may not have the wall on).

- After ANY non-status action (`install`, `upgrade`, `enable`, `disable`, `uninstall`, `purge`, `level`, `edit`, `hard-one-shot`, `manager-run`, `inline-run`), end by emitting the resolved status (run the `status` Bash block, or reuse a result you already have). At minimum print:

- For `install`/`upgrade` that NEWLY registered, and for `uninstall`/`purge`, also surface the `/reload` note.

-

- > If the wall ended up ARMED (`install`, `enable`, `hard-one-shot` while running), use the always-allowed status path described in the `status` section — the full Bash block is denied under an armed wall. Always include the literal exit command in the dump.

-

- ---

-

- ## Manager discipline

-

- This skill follows the same Manager rules it installs. For any real implementation triggered via `hard-one-shot` / `manager-run` / `inline-run`, it scans ALL available agents and delegates to the single best-matching expert — it does not write code, run builds, or hand-fix bugs itself. Orchestration only.

- | `uninstall`/`purge` requested while `state.hard` is true | Run the bare exempt disarm command as its own FIRST Bash call, then the deregistration block — never edit settings under an armed wall, and never merge the two calls. |

- | Any Bash block here denied by the guard with `Manager HARD wall is ON` | You appended something to the exempt command, or the wall is armed and you used a `BT_ROOT=`/`&& echo` block. Re-issue the bare `node <ABS_CWD>/.claude/brewtools/manager/manager-state.mjs set hard=false`, or delegate the block to a subagent. |

- | Neither the skill dir nor any cached plugin dir yields `.claude-plugin/plugin.json` | ERROR: `manager-setup: cannot locate plugin root — install/update brewtools first.` STOP. |

- | Intent ambiguous / conflicting (incl. hard-one-shot vs manager-run) | `AskUserQuestion` with candidate actions. |