ai-consultants 5.1.2 → 5.2.0

package/README.md

@@ -1,8 +1,8 @@
-# AI Consultants v5.1.2
+# AI Consultants v5.2.0
 
 > **Coverage, not a single guess.** A panel of up to 10 frontier models from different vendors fans out on your question in parallel and hands you the *union* of what they collectively see — the risks, edge cases, and approaches a single model misses.
 
-[![Version](https://img.shields.io/badge/version-5.1.2-blue.svg)](https://github.com/matteoscurati/ai-consultants)
+[![Version](https://img.shields.io/badge/version-5.2.0-blue.svg)](https://github.com/matteoscurati/ai-consultants)
 [![License](https://img.shields.io/badge/license-MIT-green.svg)](LICENSE)
 [![Claude Code Skill](https://img.shields.io/badge/Claude%20Code-Skill-orange.svg)](https://docs.anthropic.com/en/docs/claude-code/skills)
 [![GitHub stars](https://img.shields.io/github/stars/matteoscurati/ai-consultants?style=social)](https://github.com/matteoscurati/ai-consultants)
@@ -18,7 +18,8 @@
 - [Supported CLI Agents](#supported-cli-agents)
   - [Claude Code](#claude-code)
   - [OpenAI Codex CLI](#openai-codex-cli)
-  - [Gemini CLI](#gemini-cli)
+  - [Antigravity (agy)](#antigravity-agy)
+  - [Gemini CLI (legacy)](#gemini-cli-legacy)
   - [Cursor / Copilot / Windsurf](#cursor--copilot--windsurf-via-skillport)
   - [Standalone Bash](#standalone-bash)
 - [Consultants](#consultants)
@@ -100,7 +101,8 @@ Before installing AI Consultants, ensure you have the following dependencies ins
 |------------|---------|
 | **jq** | JSON processing |
 | **curl** | HTTP requests and connectivity |
-| **Bash 4.0+** | Script execution (macOS ships with 3.2) |
+| **Bash 3.2+** | Script execution (the stock macOS `/bin/bash` 3.2 is supported) |
+| **bc** | Cost arithmetic |
 
 ### Installation by Platform
 
@@ -110,21 +112,22 @@ Before installing AI Consultants, ensure you have the following dependencies ins
 # Install Homebrew if not already installed
 /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
 
-# Install required dependencies
-brew install jq bash coreutils
+# Install required dependencies (coreutils is optional: it provides
+# `timeout`; a built-in fallback is used without it)
+brew install jq coreutils
 
 # Verify installation
 jq --version && bash --version | head -1
 ```
 
-> **Note:** macOS ships with Bash 3.2. The Homebrew version (4.x) is installed to `/opt/homebrew/bin/bash`.
+> **Note:** the scripts target the stock macOS Bash 3.2 (`#!/bin/bash`); installing a newer Bash is not required.
 
 #### Linux (Ubuntu/Debian)
 
 ```bash
 # Install required dependencies
 sudo apt-get update
-sudo apt-get install -y jq curl bash
+sudo apt-get install -y jq curl bash bc
 
 # Verify installation
 jq --version && bash --version | head -1
@@ -134,7 +137,7 @@ jq --version && bash --version | head -1
 
 ```bash
 # Install required dependencies
-sudo dnf install -y jq curl bash
+sudo dnf install -y jq curl bash bc
 
 # Verify installation
 jq --version && bash --version | head -1
@@ -144,7 +147,7 @@ jq --version && bash --version | head -1
 
 ```bash
 # Install required dependencies
-sudo pacman -S jq curl bash
+sudo pacman -S jq curl bash bc
 
 # Verify installation
 jq --version && bash --version | head -1
@@ -160,7 +163,7 @@ wsl --install
 
 # After restart, open WSL and follow Linux instructions
 sudo apt-get update
-sudo apt-get install -y jq curl bash
+sudo apt-get install -y jq curl bash bc
 ```
 
 Alternatively, use **Git Bash** or **MSYS2** with the required packages.
@@ -171,7 +174,7 @@ For CLI-based consultants, you'll also need:
 
 | Dependency | Required for |
 |------------|--------------|
-| **Node.js 18+** | Gemini CLI, Codex CLI, Qwen CLI, MiniMax CLI |
+| **Node.js 18+** | Codex CLI, Qwen CLI, MiniMax CLI |
 | **Python 3.8+** | Mistral Vibe CLI |
 
 ```bash
@@ -201,7 +204,7 @@ AI Consultants follows the open [Agent Skills standard](https://agentskills.io),
 
 ### Claude Code
 
-> **Status:** ✅ Native support
+> **Status:** ✅ Native support (headless run verified with Claude Code 2.1.303 on 2026-10-19)
 
 **Installation:**
 
@@ -234,18 +237,27 @@ slash command.
 
 ### OpenAI Codex CLI
 
-> **Status:** ✅ Compatible
+> **Status:** ✅ Compatible (headless run verified with Codex CLI 0.163 on 2026-10-19)
 
 **Installation:**
 
 ```bash
 git clone https://github.com/matteoscurati/ai-consultants.git ~/.codex/skills/ai-consultants
 ~/.codex/skills/ai-consultants/scripts/doctor.sh --fix
+# Or share an existing installation through the cross-agent skills directory:
+# ln -s ~/.claude/skills/ai-consultants ~/.agents/skills/ai-consultants
 ```
 
-**Commands:**
+**Usage:**
 
-Use the same slash commands as Claude Code. Codex CLI loads skills from `~/.codex/skills/`.
+Codex has no custom slash commands: it loads the skill from `~/.codex/skills/`
+or `~/.agents/skills/` and runs it when asked, for example *"use the
+ai-consultants skill to review src/auth.ts"*. The consultation calls other CLIs
+and APIs over the network and writes under `~/.cache` and `~/.config`, so
+Codex's default sandbox cannot run it: approve the escalation when Codex asks,
+or start Codex with `--sandbox danger-full-access`. Codex's automatic approval
+review (`--approve-for-me`) rejects that escalation because it sends your files
+to external providers, so the approval has to come from you.
 
 **Self-Exclusion:** Codex consultant is automatically excluded when invoked from Codex CLI.
 
@@ -257,28 +269,40 @@ Use the same slash commands as Claude Code. Codex CLI loads skills from `~/.code
 
 ---
 
-### Gemini CLI
+### Antigravity (agy)
 
-> **Status:** ✅ Compatible
+> **Status:** ✅ Compatible (headless run verified with agy 1.4.2 on 2026-10-19)
+
+Antigravity (`agy`) replaced the Gemini CLI for individual Google accounts.
 
 **Installation:**
 
 ```bash
-git clone https://github.com/matteoscurati/ai-consultants.git ~/.gemini/skills/ai-consultants
-~/.gemini/skills/ai-consultants/scripts/doctor.sh --fix
+# Per project (agy reads <project>/.agents/skills/)
+mkdir -p .agents/skills
+git clone https://github.com/matteoscurati/ai-consultants.git .agents/skills/ai-consultants
+# Or for every project: agy's user skill directory is ~/.gemini/config/skills/
 ```
 
-**Commands:**
+**Usage:** ask in natural language, for example *"use the ai-consultants skill
+to compare these two designs"*. agy runs the consultation as a background task
+and follows its progress; in print mode (`agy -p`) it waits at most 30 minutes
+for background tasks.
 
-Use the same slash commands as Claude Code. Gemini CLI loads skills from `~/.gemini/skills/`.
+**Self-Exclusion:** the Gemini consultant is excluded when invoked from agy
+(`INVOKING_AGENT=gemini`).
 
-**Self-Exclusion:** Gemini consultant is automatically excluded when invoked from Gemini CLI.
+---
 
-**Verify:**
+### Gemini CLI (legacy)
 
-```bash
-~/.gemini/skills/ai-consultants/scripts/doctor.sh
-```
+> **Status:** ⚠️ Legacy. Google stopped serving the Gemini CLI to individual
+> accounts on 2026-06-18; use [Antigravity (agy)](#antigravity-agy) instead.
+
+Accounts that can still use it get the skill from `~/.gemini/skills/` or
+`~/.agents/skills/`, through natural language: the Gemini CLI loads only
+`.toml` custom commands, so the `/ai-consultants:*` slash commands are not
+available there. The Gemini consultant is excluded as for agy.
 
 ---
 
@@ -416,7 +440,8 @@ the observed version is response provenance only.
 
 ### Installing Consultant CLIs
 
-At least 2 consultant CLIs are required:
+At least 2 consultants are required (a CLI, or an API key for API mode). The
+first three below make up the `balanced` preset:
 
 ```bash
 curl -fsSL https://antigravity.google/cli/install.sh | bash  # Gemini (Antigravity CLI: agy)
@@ -462,11 +487,12 @@ accepted reasoning effort: Grok Build uses `xhigh`, while GLM and DeepSeek use
 effort silently.
 
 The maximum preset also applies the transport budgets proven by the live panel
-smoke: four bounded advisory turns for Mistral/Grok, a 600-second Token Plan
-window for Qwen3.8-Max, a 600-second DeepSeek reasoning window, and larger
-OpenAI-compatible output budgets for Qwen, GLM, and DeepSeek. MiniMax M3
-likewise receives a 16,384-token completion budget in the maximum tier and a
-provider-specific compact Markdown contract through mmx's native system channel.
+smoke: four bounded advisory turns for Mistral/Grok, the maximum-tier Qwen3.8-Max
+and DeepSeek timeouts (`QWEN3_MAX_QUALITY_TIMEOUT`/`DEEPSEEK_MAX_QUALITY_TIMEOUT`,
+default 0: no limit), and larger OpenAI-compatible output budgets for Qwen, GLM,
+and DeepSeek. MiniMax M3 likewise receives a 16,384-token completion budget in
+the maximum tier and the shared JSON response contract through mmx's native
+system channel.
 Provider output is tagged `structured`, `fallback`, or `error`;
 malformed/truncated JSON fails
 closed. In coverage/union mode, only locally atomized `summary`, `pros`,
@@ -484,30 +510,40 @@ and cannot support a comprehensive coverage claim.
 | Consultant | `max_quality` | Premium | Standard | Economy |
 |------------|---------------|---------|----------|---------|
 | Claude | claude-fable-5-1 | claude-fable-5-1 | claude-opus-5-5 | claude-haiku-4-5 |
-| Gemini CLI | Gemini 3.8 Flash (High) | Gemini 3.8 Flash (High) | Gemini 3.8 Flash (High) | Gemini 3.8 Flash (Low) |
-| Gemini API | gemini-3.1-pro-preview | gemini-3.1-pro-preview | gemini-3.1-pro-preview | gemini-3.1-pro-preview |
-| Codex | gpt-6-astra | gpt-6-astra | gpt-5.6-terra | gpt-5.6-luna |
-| Mistral CLI | mistral-medium-3.5 | mistral-medium-3.5 | mistral-medium-3.5 | devstral-small-2 |
-| Mistral API | mistral-large-3 | mistral-large-3 | mistral-large-3 | mistral-large-3 |
+| Gemini CLI | Gemini 4 Argon (High) | Gemini 4 Argon (High) | Gemini 3.8 Flash (High) | Gemini 3.8 Flash (Low) |
+| Gemini API | gemini-4-argon | gemini-4-argon | gemini-3.1-pro-preview | gemini-3.1-pro-preview |
+| Codex | gpt-6-astra | gpt-6.1-sol | gpt-5.6-terra | gpt-5.6-luna |
+| Mistral CLI | mistral-medium-3.5 | mistral-medium-3.5 | mistral-medium-3.5 | mistral-medium-3.5 |
+| Mistral API | mistral-large-4 | mistral-large-4 | mistral-large-4 | mistral-large-4 |
 | DeepSeek | deepseek-flash | deepseek-flash | deepseek-flash | deepseek-flash |
-| GLM | glm-5.3-flash | glm-5.3-flash | glm-5.3-flash | glm-4-flash |
+| GLM | glm-5.3 | glm-5.3-flash | glm-5.3-flash | glm-4-flash |
 | Grok | grok-4.7 | grok-4.7 | grok-4.5 | grok-4.5 |
-| Qwen3 | qwen3.8-max when Token Plan is configured; otherwise qwen3.7-max | qwen3.7-max | qwen3.6-35b-a3b | qwen3-32b |
+| Qwen3 | qwen3.8-max | qwen3.8-max | qwen3.8-flash | qwen3.8-flash |
 | Kimi | kimi-code/k3-256k | kimi-code/k3 | kimi-code/k3 | kimi-code/k3 |
 | MiniMax | MiniMax-M3 | MiniMax-M2.7 | MiniMax-M2.7 | MiniMax-M2.5 |
 
-Promotion is transport-specific. Gemini 3.8 Flash High is the new CLI target;
-Low serves economy. Both are available in the installed `agy models` inventory;
-High/Low CLI smokes passed on 2026-09-28 with capability-probed identity. Earlier Gemini 3.7 evidence remains historical.
-The Google API default remains `gemini-3.1-pro-preview`; `gemini-3.8-flash` is
-an explicit API opt-in whose live transport is not yet verified. The new Mistral API IDs (`mistral-medium-3-5`,
-`mistral-large-2512`, `mistral-small-2603`) remain catalogued opt-ins. Claude
+Promotion is transport-specific. Gemini 4 Argon (GA 2026-10-13) is the Gemini
+default and maximum target: `Gemini 4 Argon (High)` on `agy` and `gemini-4-argon`
+on the API. An authorized `agy` consultation smoke returned a structured answer
+with capability-probed identity on 2026-10-16; the API transport is covered by
+offline tests only. Google excludes Argon from the Free tier and the AI Pro
+plan, so the `agy` listing alone is not an access check. Gemini 3.8 Flash keeps
+the CLI standard (High) and economy (Low) tiers; the API standard and economy
+tiers keep `gemini-3.1-pro-preview`, and `gemini-3.8-flash` remains an explicit
+API opt-in whose live transport is not yet verified. Mistral Large 4 (`mistral-large-4`, public preview) is the
+Mistral API target for every tier. An authorized consultation smoke returned a
+structured answer with provider-reported identity on 2026-10-15; the preview is absent from
+`GET /v1/models`, so that listing is not an access check. Vibe keeps
+`mistral-medium-3.5`. The other Mistral
+API IDs (`mistral-medium-3-5`, `mistral-large-2512`, `mistral-small-2603`)
+remain catalogued opt-ins. Claude
 Fable 5.1 is the default/premium/maximum target; use
 `CLAUDE_MODEL=claude-opus-5-5` for the lower-cost standard override. Selecting a preset later intentionally reapplies that
 preset's tier and can replace an explicit model override for the run.
-Qwen3.8-Max is likewise selected by `max_quality` only when API mode already
-points at an authenticated OpenAI-compatible Token Plan `/chat/completions`
-endpoint; the preset never repoints a DashScope key or CLI installation.
+Qwen3.8-Max serves the default, premium and maximum tiers on every Qwen
+transport (the `qwen` CLI, the Token Plan and DashScope); `max_quality` adds
+`xhigh` effort only on the OpenAI-compatible wire and never repoints a key or
+endpoint.
 
 ### Usage
 
@@ -611,11 +647,13 @@ later run to adapt when a CLI or credential changes. Environment variables,
 `--set`, and unmarked values remain explicit user choices, except for the exact
 historical generated Claude default described below.
 Managed model defaults use `# ai-consultants:default`; `configure` upgrades
-un-pinned historical `CLAUDE_MODEL=claude-opus-4-8` and `claude-opus-5` defaults
-to Fable 5.1, along with the exact managed Gemini CLI, GLM, and Grok defaults.
-To keep the lower-cost Opus 5 choice, run
-`ai-consultants configure --set CLAUDE_MODEL=claude-opus-5`; explicit model
-overrides are stored with `# ai-consultants:pin`.
+the exact historical defaults it generated when they are not pinned (for
+example `CLAUDE_MODEL=claude-opus-4-8`/`claude-opus-5` to Fable 5.1, and the
+previous Gemini, Codex, Mistral API, Qwen, GLM, Grok and DeepSeek defaults),
+and turns the fixed timeouts it generated through 5.1.2 into 0 (no limit).
+To keep a model, run for example
+`ai-consultants configure --set CLAUDE_MODEL=claude-opus-5-5`; explicit model
+overrides are stored with `# ai-consultants:pin` and never migrated.
 
 Enter credentials through `--interactive`/`--advanced` or export them before the
 run; avoid passing API keys through `--set`, where the shell may retain them in
@@ -634,7 +672,7 @@ MAX_SESSION_COST=1.00        # USD budget cap (paired with ENABLE_BUDGET_LIMIT=t
 KIMI_MODEL=kimi-code/k3      # Pin the Kimi consultant to K3
 CLAUDE_API_MAX_TOKENS=16384  # Shared thinking + visible-output budget in API mode
 MISTRAL_CLI_MODEL=mistral-medium-3.5 # Vibe alias; MISTRAL_MODEL remains API-only
-GEMINI_MODEL="Gemini 3.8 Flash (High)" # agy CLI; API uses GEMINI_API_MODEL
+GEMINI_MODEL="Gemini 4 Argon (High)" # agy CLI; API uses GEMINI_API_MODEL
 ```
 
 Full reference: [`references/configuration.md`](references/configuration.md). Copy-paste workflows: [`docs/RECIPES.md`](docs/RECIPES.md). For category-aware preset suggestions: `ai-consultants doctor --suggest-preset --question "..."`.
@@ -713,6 +751,25 @@ Classify -> Route -> Fan out (parallel) -> Coverage synthesis
    non-comprehensive. Override with `--strategy compare_only` (side-by-side) or
    `majority` (a single blended recommendation).
 
+### Waiting and stopping
+
+Consultants and synthesis have **no time limit by default**: reasoning models
+can take minutes, and a call cut off by a client timeout is usually still
+billed while its answer is lost. Whoever started the run decides when to stop
+waiting. The run logs `Output: <dir>` near the start and
+`Still waiting after Ns for: <names>` every 30 seconds, and accepts:
+
+- `touch <dir>/.proceed` (or `kill -USR1 <pid>`, or a first Ctrl-C): continue
+  with the answers already received. Consultants still running are stopped and
+  reported as `stopped`; sent during synthesis, it skips the synthesis.
+- `touch <dir>/.abort` (or SIGTERM, or a second Ctrl-C): end the run and keep
+  the partial output.
+
+A stopped call may still be billed for what it generated. Never stop a run with
+SIGKILL: the shell cannot clean up, so consultant CLIs would keep running. A
+positive `<NAME>_TIMEOUT` restores a hard limit for that consultant; see
+[`references/configuration.md`](references/configuration.md#timeouts-waiting-and-retries).
+
 ### Output
 
 Each consultation generates:

package/SKILL.md

@@ -5,10 +5,10 @@ license: MIT
 compatibility: Requires bash, jq, and at least 2 AI CLI tools (agy, codex, vibe, etc.). macOS and Linux.
 metadata:
   author: matteoscurati
-  version: 5.1.2
+  version: 5.2.0
 ---
 
-# AI Consultants v5.1.2 - AI Expert Panel
+# AI Consultants v5.2.0 - AI Expert Panel
 
 **Coverage, not a single guess.** Convene a panel of AI "consultants" from different vendors for coding questions: they fan out in parallel and you get the *union* of what they collectively see — the risks, edge cases, and approaches a single model misses. Each consultant has a **configurable persona** that decorrelates its analysis.
 
@@ -25,15 +25,20 @@ only telling the user to type the slash command. Route through
 `scripts/consult_all.sh`; do not call provider adapters individually.
 
 1. Identify the invoking host and set `INVOKING_AGENT` explicitly:
-   `claude` for Claude Code, `codex` for Codex CLI, or `gemini` for Gemini CLI.
+   `claude` for Claude Code, `codex` for Codex CLI, or `gemini` for
+   Antigravity (`agy`) and the legacy Gemini CLI.
    This is the boundary that prevents the host from consulting or synthesizing
    with itself.
-2. Resolve the skill root from `AI_CONSULTANTS_DIR` when set. Otherwise use the
-   host installation (`~/.claude/skills/ai-consultants` for Claude Code,
-   `~/.codex/skills/ai-consultants` for Codex, or
-   `~/.gemini/skills/ai-consultants` for Gemini).
-3. Preserve consultation options such as `--preset` and `--strategy` as CLI
-   arguments before the question. Pass referenced files as paths, optionally
+2. Resolve the skill root from `AI_CONSULTANTS_DIR` when set. Otherwise the
+   skill root is the directory that contains this `SKILL.md` (for example
+   `~/.claude/skills/ai-consultants`, `~/.agents/skills/ai-consultants` or
+   `<project>/.agents/skills/ai-consultants`); run everything from there.
+   The consultation calls external CLIs and APIs over the network and writes
+   under `~/.cache`, `~/.config` and the CLIs' own directories, so a sandboxed
+   shell (for example Codex's default sandbox) must run it with network access
+   and those writes allowed, or ask the user for that permission.
+3. Preserve consultation options such as `--mode`, `--preset` and `--strategy`
+   as CLI arguments before the question. Pass referenced files as paths, optionally
    suffixed with `@PRIMARY` or `@CONTEXT`; do not paste their contents. Capture
    the caller's project directory before entering the skill root and pass it as
    `--context-root`, so the runtime can privately stage only regular in-project
@@ -43,15 +48,28 @@ only telling the user to type the slash command. Route through
    question. Use `--query-file`; never pass raw user text as a shell argument or
    through an unquoted heredoc. Shell-quote every context-file path as one
    argument, and install a shell trap that removes only that exact temporary
-   question file. Capture the last stdout line as the output directory, then
+   question file (some harnesses, such as Codex, reject `rm -f`; `unlink` works
+   there). Capture the last stdout line as the output directory, then
    read `report.md` and
    `synthesis.json` (or the individual consultant JSON files when synthesis is
    unavailable).
-5. Keep `INVOKING_AGENT` and the private `--query-file` handoff on follow-ups.
+5. There is no generation time limit by default, and you decide when to stop
+   waiting. Run the consultation in the background when your shell tool allows
+   it, and read its log: it prints `Output: <dir>` near the start and
+   `Still waiting after Ns for: <names>` every 30 seconds. `touch <dir>/.proceed`
+   continues with the answers received (consultants still running are stopped;
+   during synthesis it skips the synthesis); `touch <dir>/.abort` ends the run.
+   A stopped call may still be billed, so keep waiting unless the user wants
+   results now or a consultant looks stuck. Never stop the run with SIGKILL.
+   Your own harness may cap the wait (for example, `agy` in print mode waits at
+   most 30 minutes for background tasks); touch `.proceed` before such a cap
+   rather than letting the harness kill the run.
+6. Keep `INVOKING_AGENT` and the private `--query-file` handoff on follow-ups.
    A targeted follow-up to the invoking consultant must fail closed. For a
-   Claude invocation, `claude.json` must not be produced and Claude must not be
-   the synthesis provider. Treat either condition as a self-exclusion failure,
-   not a valid consultation.
+   Claude invocation, `claude.json` must not be produced and the
+   `synthesis_provider` field of `synthesis.json` must not be `claude` (the same
+   holds for `codex` and `gemini` hosts). Treat either condition as a
+   self-exclusion failure, not a valid consultation.
 
 Claude Code example:
 
@@ -102,20 +120,22 @@ model, and structured-output surface. Compatible version changes are accepted;
 the observed version is recorded only as response provenance.
 
 The `max_quality` preset enables all 10 consultants. It uses the separately
-smoke-tested K3-256k and MiniMax M3 targets, plus Qwen3.8-Max when an authenticated Token Plan transport is
-already configured. Gemini 3.8 Flash High is the new `agy` target for maximum,
-premium, and standard CLI tiers; Low serves economy. The installed model inventory
-lists both; High/Low CLI smokes passed on 2026-09-28 with capability-probed identity.
-The Gemini API default remains 3.1 Pro; `gemini-3.8-flash` is an explicit API opt-in. Claude Fable 5.1 is the default/premium/maximum target; the new
-Mistral API IDs remain explicit opt-ins.
+smoke-tested K3-256k and MiniMax M3 targets, GLM-5.3 (the flagship; Flash serves
+premium/standard) and Qwen3.8-Max (also the Qwen default; Flash serves
+standard/economy). Gemini 4 Argon (High) is the `agy` target for the default
+and maximum tiers (`gemini-4-argon` on the API); one authorized `agy` smoke passed on
+2026-10-16 with capability-probed identity, and the API transport is offline-tested only.
+Gemini 3.8 Flash High/Low keep the CLI standard/economy tiers, the API standard/economy
+tiers keep 3.1 Pro, and `gemini-3.8-flash` is an explicit API opt-in. Claude Fable 5.1 is the default/premium/maximum target. Mistral
+API mode uses Mistral Large 4 in every tier; the other new Mistral API IDs remain explicit opt-ins.
 Grok, GLM, and DeepSeek receive their
 highest accepted reasoning effort in this preset (`xhigh`, `max`, and `max`
 respectively); no failed target is silently replaced.
 
 Maximum-quality transports use smoke-tested runtime budgets: four advisory
-turns for Mistral/Grok, extended Qwen/DeepSeek timeouts, and larger response
-budgets for reasoning APIs and a compact native-system Markdown contract for
-MiniMax M3. Responses carry an explicit `structured`,
+turns for Mistral/Grok, no generation time limit, and larger response
+budgets for reasoning APIs; MiniMax M3 receives the shared JSON contract in
+its native system channel. Responses carry an explicit `structured`,
 `fallback`, or `error` quality. Truncated JSON fails closed. Coverage/union
 synthesis source-attributed claims use only locally normalized atomic fields;
 fallback prose remains visible as context-only manual-review material. Inspect
@@ -224,7 +244,7 @@ Use `ai-consultants configure --show-parameters` for the exact accepted keys,
 `--set KEY=VALUE` for repeatable automation, or `--advanced` to review all of
 them interactively.
 
-CLI/API switching: `GEMINI_USE_API`, `CODEX_USE_API`, `CLAUDE_USE_API`, `MISTRAL_USE_API`, `QWEN3_USE_API`.
+CLI/API switching: `GEMINI_USE_API`, `CODEX_USE_API`, `CLAUDE_USE_API`, `MISTRAL_USE_API`, `QWEN3_USE_API`, `GROK_USE_API`, `MINIMAX_USE_API`.
 
 See [Full Configuration Reference](references/configuration.md) for all variables, model overrides, tiers, timeouts, and optimization settings.
 

package/bin/ai-consultants

@@ -191,6 +191,32 @@ EOF
   install)
     COMMANDS_SRC="$PROJECT_ROOT/.claude/commands"
     COMMANDS_DST="$HOME/.claude/commands"
+
+    # The commands cd into ${AI_CONSULTANTS_DIR:-$HOME/.claude/skills/ai-consultants}.
+    # A curl | bash install lives there and outlasts any npx cache, so when a
+    # complete one exists, install its own commands: they stay matched to the
+    # scripts they run. Only without one do the commands point at this package.
+    DEFAULT_SKILL_ROOT='${AI_CONSULTANTS_DIR:-$HOME/.claude/skills/ai-consultants}'
+    default_root="$HOME/.claude/skills/ai-consultants"
+    default_root_resolved=$(cd -P "$default_root" 2>/dev/null && pwd || true)
+    use_default_root=false
+    if [ -n "$default_root_resolved" ] && [ "$PROJECT_ROOT" != "$default_root_resolved" ] \
+      && [ -x "$default_root/scripts/consult_all.sh" ] \
+      && [ -x "$default_root/scripts/doctor.sh" ] \
+      && [ -x "$default_root/scripts/followup.sh" ] \
+      && [ -d "$default_root/.claude/commands" ] \
+      && grep -q '^name: ai-consultants$' "$default_root/SKILL.md" 2>/dev/null; then
+      use_default_root=true
+      COMMANDS_SRC="$default_root/.claude/commands"
+      default_version=$(grep -E '^AI_CONSULTANTS_VERSION=' "$default_root/scripts/config.sh" 2>/dev/null \
+        | head -1 | sed -E 's/.*"([^"]+)".*/\1/' || true)
+      echo "Using the existing installation at $default_root (v${default_version:-unknown})"
+      if [ "$default_version" != "$VERSION" ]; then
+        echo "Note: this package is v$VERSION. To upgrade that installation run:" >&2
+        echo "      bash \"$default_root/scripts/install.sh\" --update" >&2
+      fi
+    fi
+
     if [ ! -d "$COMMANDS_SRC" ]; then
       echo "Error: slash commands directory not found at $COMMANDS_SRC" >&2
       exit 1
@@ -215,10 +241,45 @@ EOF
       echo "Pruned $pruned slash command(s) removed in a previous release"
     fi
 
+    # Without a usable default install, point the default at this package
+    # root so the installed commands can run.
+    rewrite_root=""
+    if [ "$use_default_root" = false ] && [ "$PROJECT_ROOT" != "$default_root_resolved" ]; then
+      # The root lands inside "${AI_CONSULTANTS_DIR:-<root>}", where Bash also
+      # treats a single quote as quoting: an unmatched one is a syntax error.
+      case "$PROJECT_ROOT" in
+        *'"'*|*"'"*|*'$'*|*'`'*|*'\'*)
+          echo "Warning: package path contains shell metacharacters; set AI_CONSULTANTS_DIR=$PROJECT_ROOT before using the slash commands" >&2
+          ;;
+        *)
+          rewrite_root="$PROJECT_ROOT"
+          ;;
+      esac
+      case "$PROJECT_ROOT" in
+        */_npx/*)
+          echo "Warning: this package runs from the npx cache ($PROJECT_ROOT), which npm may delete." >&2
+          echo "         For stable slash commands run: npm install -g ai-consultants && ai-consultants install" >&2
+          ;;
+      esac
+    fi
+
     copied=0
     for f in "$COMMANDS_SRC"/ai-consultants:*.md; do
       [ -f "$f" ] || continue
-      cp "$f" "$COMMANDS_DST/"
+      dest="$COMMANDS_DST/$(basename "$f")"
+      if [ -n "$rewrite_root" ]; then
+        # awk index/substr: a literal replacement, independent of Bash
+        # version quirks in ${var//pattern/replacement} (e.g. "&" in 5.2).
+        awk -v old="$DEFAULT_SKILL_ROOT" -v new="\${AI_CONSULTANTS_DIR:-$rewrite_root}" '
+          { out = ""; line = $0
+            while ((i = index(line, old)) > 0) {
+              out = out substr(line, 1, i - 1) new
+              line = substr(line, i + length(old))
+            }
+            print out line }' "$f" > "$dest"
+      else
+        cp "$f" "$dest"
+      fi
       copied=$((copied + 1))
     done
     if [ "$copied" -eq 0 ]; then