ai-consultants 4.0.1 → 4.0.3

package/.env.example

@@ -188,13 +188,17 @@ QWEN3_CMD=qwen
 # is installed; the xAI API is selected when the CLI is unavailable and a key
 # exists. A failed CLI launch also falls back to the API at runtime.
 # GROK_USE_API=false
+# Grok OAuth concurrency: shared keeps one runner-owned credential generation
+# so parallel consultants can safely reuse refresh rotation. serialized holds
+# the adapter lock for the full CLI call and is intended for diagnostics.
+# GROK_OAUTH_MODE=shared
 
 # =============================================================================
 # API-ONLY AND FALLBACK CONFIGURATION
 # =============================================================================
 
 # GLM - The Methodologist
-GLM_MODEL=glm-5.3
+GLM_MODEL=glm-5.3-flash
 GLM_TIMEOUT=180
 GLM_API_MAX_TOKENS=16384
 GLM_API_URL=https://api.z.ai/api/coding/paas/v4/chat/completions
@@ -205,6 +209,7 @@ GROK_CMD=grok
 GROK_MODEL=grok-4.6
 GROK_TIMEOUT=180
 GROK_MAX_TURNS=4
+GROK_OAUTH_MODE=shared
 GROK_API_MAX_TOKENS=4096
 GROK_API_URL=https://api.x.ai/v1/chat/completions
 # GROK_REASONING_EFFORT= # max_quality sets xhigh (Grok Build maximum)

package/CLAUDE.md

@@ -6,7 +6,7 @@ AI Consultants is a multi-model coverage system that queries up to 10 AI consult
 
 **Self-Exclusion**: The invoking agent is automatically excluded from both the panel and synthesis. Claude Code won't query or synthesize with Claude, Codex CLI won't query or synthesize with Codex, etc.
 
-**Version**: 4.0.1
+**Version**: 4.0.3
 
 ## Distribution
 
@@ -393,7 +393,7 @@ All consultants now use premium models by default:
 | Codex | gpt-5.6-sol |
 | Mistral | CLI: mistral-medium-3.5; API: mistral-large-3 |
 | DeepSeek | deepseek-v4-pro |
-| GLM | glm-5.3 |
+| GLM | glm-5.3-flash |
 | Grok | grok-4.6 |
 | Qwen3 | qwen3.7-max |
 | Kimi | kimi-code/k3 |
@@ -760,6 +760,27 @@ curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/s
 
 ## Changelog
 
+### v4.0.3
+
+- **GLM's active catalog target moves from `glm-5.3` to `glm-5.3-flash` without changing transport or tier semantics.** `scripts/config.sh:189` changes the clean-install default, and the maximum/premium/standard branches at `scripts/config.sh:626-656` select the same exact provider ID. `max_quality` still supplies `GLM_REASONING_EFFORT=max`; economy deliberately remains `glm-4-flash`.
+- **Existing managed installs advance without rewriting explicit choices.** `scripts/configure.sh:310-318` migrates the two exact historical generated defaults, `glm-5.2` and `glm-5.3`, to Flash and records `# ai-consultants:default`. A `# ai-consultants:pin` survives, including an explicit old `glm-5.3`; `scripts/test_configure.sh:449-461` covers both current-default migration and pin preservation.
+- **Identity and pricing stay honest.** The OpenAI-compatible adapter sends `glm-5.3-flash`, `reasoning_effort=max`, and the 16,384-token GLM budget; the synthetic transport test requires provider-reported identity and requested-model metadata (`scripts/test_api_transport.sh:253-263`). `docs/cost_rates.json` marks Flash unpriced and retains a zero/unpriced entry for legacy `glm-5.3` pins, avoiding the generic fallback rate.
+- **The exact ai-consultants transport completed one authorized no-retry live smoke.** Z.AI returned HTTP 200 and provider-reported `glm-5.3-flash` in about 12 seconds with 37 measured input tokens and 308 output tokens. The provider accepted max effort but returned near-structured text rather than the requested JSON; the adapter preserved `PONG` through its explicit unstructured fallback with confidence 5, so strict schema adherence is not claimed. PR #15 CI, the post-bump 23-suite gate, 452 core checks, configure 94/94, API transport 88/88, and ShellCheck are green.
+
+  Deliberately unchanged: `delegation-kit` remained read-only; its Claude-to-Z.AI gate and runner were not copied into this API-only consultant. Installation does not silently edit personal configuration, so an existing unpinned `.env` advances when the user runs `configure`; explicit pins remain user-owned. The separate `agent/model-defaults-4.1.0-wip` branch is not part of this patch release.
+
+### v4.0.2
+
+- **The Grok adapter used to copy a rotating OAuth credential into a disposable HOME and delete any refreshed token at exit.** Parallel calls amplified the race: both could seed from one refresh token, one rotation vanished with its temp directory, and the ambient session could be left stale. `scripts/lib/grok_oauth.sh` now owns a durable generation under the XDG data root and `scripts/query_grok.sh` defaults to `GROK_OAUTH_MODE=shared`.
+- **Shared means credential state only.** Every invocation still has a private HOME, workspace, prompt file, output, capability-probe HOME, permission mode, strict sandbox, empty tool surface, model, and effort. Concurrent processes use one runner-owned `GROK_HOME`, allowing Grok Build's native auth lock to coordinate refresh while inference remains outside the ai-consultants lock.
+- **Publication is optimistic and atomic.** A recoverable directory lock bounds seed, bootstrap, inventory, and publication; valid rotation is staged mode 600 and renamed into the ambient `auth.json`. The marker binds generation and last-synced SHA-256. If a concurrent external `grok login` changes the ambient digest, that login wins, the consultation fails temporarily, and no API fallback occurs.
+- **Malformed and obsolete state fails closed but recovers forward.** Corrupt credentials never replace ambient auth; dead locks are reclaimed; unpersistible refreshes publish no answer. A later run reseeds from valid ambient state. Generations carrying the exact legacy runner-owned `config.toml` signature are superseded without modification or deletion.
+- **Grok 1.0.4 required exact-transport fixes beyond the reference implementation.** Its pristine-home bootstrap lazily creates metadata/cache/DB state; capability probes must use a separate credential-free HOME; authenticated inventory is bounded and coordinated. Provider-owned `config.toml` cannot be replaced before each inventory, and the display sentence “You are logged in with grok.com.” is not stable enough to be an auth gate. Exit 0, a still-valid credential, and the requested inventory model are the attestation.
+- **Secrets remain private.** The prompt and tokens never enter argv. Grok dispatch uses redacted `run_query` errors, so token-shaped provider stderr is absent from stdout, normal logs, error envelopes, and artifacts. Forced API mode remains stateless; missing/auth-unavailable CLI fallback and post-dispatch fallback suppression are unchanged.
+- **Verification closes both concurrency layers.** The adapter suite grows to 88 checks and covers simultaneous fake dispatch, rotation/reuse, external-login CAS, corrupt state, dead locks, atomic policy reconciliation, legacy migration, stateless API mode, serialized mode, and token leakage. A real two-process `grok-4.6`/`xhigh` smoke returned structured `PONG` twice with exit 0, no workspace changes, no API fallback, and fully aligned ambient/generation/marker state. Full gate: 23/23 suites, 452/452 core checks, ShellCheck and PR #14 CI green.
+
+  Deliberate residuals: superseded generations are retained because deleting one that an active process still references would be unsafe. Real forced refresh rotation and a real concurrent `grok login` were not induced; those conflict paths are verified with controlled fakes. The separate local 4.1.0 model-default WIP is not part of this release.
+
 ### v4.0.1
 
 - **The installer's first line of user-visible output was six major/minor generations stale.** `scripts/install.sh:53` printed `AI Consultants v2.10 - Installation` even while the public binary, package metadata, npm registry, active checkout, and doctor all reported v4.0.0. That did not change installed bytes or runtime behavior, but it made a successful v4 install look wrong at the exact moment users were deciding whether to trust it.

package/README.md

@@ -1,8 +1,8 @@
-# AI Consultants v4.0.1
+# AI Consultants v4.0.3
 
 > **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-4.0.1-blue.svg)](https://github.com/matteoscurati/ai-consultants)
+[![Version](https://img.shields.io/badge/version-4.0.3-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)
@@ -389,7 +389,16 @@ Grok uses Grok Build with `grok-4.6` in an isolated, tool-free sandbox. Prompts
 are passed through a private file rather than process arguments. It falls back
 to the xAI API only when the CLI is missing, cannot launch, or has no usable
 authentication and `GROK_API_KEY` is configured; post-launch request failures
-are surfaced without a silent API charge. Qwen3 and MiniMax can also switch
+are surfaced without a silent API charge. Grok CLI OAuth defaults to concurrent
+`shared` mode: HOME, workspace, prompt, output, permissions, and agent state
+remain isolated per invocation, while processes share one runner-owned
+persistent `GROK_HOME` so the CLI's own auth lock can coordinate refresh.
+Credential adoption/publication uses a short lock and digest CAS, so an
+external `grok login` wins and the affected run fails without overwriting it or
+falling back to the API. Set `GROK_OAUTH_MODE=serialized` only for diagnostic
+full-run serialization. A new generation performs one locked inventory
+bootstrap because Grok 1.0.4 lazily initializes home metadata; inference is
+never serialized by that bootstrap. Qwen3 and MiniMax can also switch
 from their CLI to API transport. Gemini, Codex, Claude, and Mistral are
 CLI/API switchable. Gemini, Grok, and Kimi verify the requested model
 against the CLI inventory before dispatch when that inventory exists. Grok and
@@ -402,7 +411,7 @@ the observed version is response provenance only.
 
 | Consultant | Default Model | Persona | Focus |
 |------------|---------------|---------|-------|
-| **GLM** | glm-5.3 | The Methodologist | Structured approaches |
+| **GLM** | glm-5.3-flash | The Methodologist | Structured approaches |
 | **DeepSeek** | deepseek-v4-pro | The Code Specialist | Algorithms, code generation |
 
 ### Installing Consultant CLIs
@@ -474,7 +483,7 @@ detail instead of reducing every consultant to its summary.
 | 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 |
 | DeepSeek | deepseek-v4-pro | deepseek-v4-pro | deepseek-v4-flash | deepseek-v4-flash |
-| GLM | glm-5.3 | glm-5.3 | glm-5.3 | glm-4-flash |
+| GLM | glm-5.3-flash | glm-5.3-flash | glm-5.3-flash | glm-4-flash |
 | Grok | grok-4.6 | grok-4.6 | 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 |
 | Kimi | kimi-code/k3-256k | kimi-code/k3 | kimi-code/k3 | kimi-code/k3 |

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: 4.0.1
+  version: 4.0.3
 ---
 
-# AI Consultants v4.0.1 - AI Expert Panel
+# AI Consultants v4.0.3 - 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.
 
@@ -88,6 +88,14 @@ Configuration (presets, strategies, features, personas, API keys) can be managed
 
 **CLI/API Mode**: Gemini, Codex, Claude, Mistral, Qwen, Grok, and MiniMax can switch between CLI and API mode via `*_USE_API` environment variables. Grok is CLI-first on Grok Build with `grok-4.6` and falls back to the xAI API when the CLI cannot run. Gemini auto-selects API mode when `GEMINI_API_KEY` is set (no `agy` install needed) and the CLI otherwise. Mistral keeps separate API (`MISTRAL_MODEL`) and Vibe (`MISTRAL_CLI_MODEL`) identifiers.
 
+Grok CLI OAuth defaults to `GROK_OAUTH_MODE=shared`: concurrent invocations
+share one runner-owned persistent `GROK_HOME` so the vendor auth lock can
+coordinate refresh, while HOME, workspace, prompt, output, permissions, and
+agent state remain isolated per invocation. A short adapter lock publishes a
+valid rotated credential atomically; an external `grok login` wins the CAS and
+makes the affected run fail without API fallback. `serialized` remains an
+explicit diagnostic mode. Forced API mode stays stateless.
+
 Grok and Kimi are compatible by capability rather than by CLI version. Before
 dispatch, their adapters verify the required headless arguments, requested
 model, and structured-output surface. Compatible version changes are accepted;

package/docs/COST_RATES.md

@@ -1,4 +1,4 @@
-# Cost Rates - AI Consultants v4.0.1
+# Cost Rates - AI Consultants v4.0.3
 
 This page documents the per-token rates used by the cost tracking system.
 
@@ -45,7 +45,8 @@ models programmatically.
 | gemini-3-flash | Legacy | $0.0005 | Removed Cursor consultant |
 | deepseek-v4-pro | Premium | $0.000435 | DeepSeek |
 | deepseek-v4-flash | Standard/Economy | $0.00014 | DeepSeek |
-| glm-5.3 | Maximum/Premium/Standard | Unpriced | GLM coding-plan endpoint |
+| glm-5.3-flash | Maximum/Premium/Standard | Unpriced | GLM coding-plan endpoint |
+| glm-5.3 | Legacy pins | Unpriced | GLM coding-plan endpoint |
 | glm-4-flash | Economy | $0.001 | GLM |
 | grok-4.6 | Maximum/Premium | $0.002 | Grok |
 | grok-4.5 | Standard/Economy | $0.002 | Grok |
@@ -86,7 +87,8 @@ models programmatically.
 | gemini-3-flash | Legacy | $0.003 | Removed Cursor consultant |
 | deepseek-v4-pro | Premium | $0.00087 | DeepSeek |
 | deepseek-v4-flash | Standard/Economy | $0.00028 | DeepSeek |
-| glm-5.3 | Maximum/Premium/Standard | Unpriced | GLM coding-plan endpoint |
+| glm-5.3-flash | Maximum/Premium/Standard | Unpriced | GLM coding-plan endpoint |
+| glm-5.3 | Legacy pins | Unpriced | GLM coding-plan endpoint |
 | glm-4-flash | Economy | $0.003 | GLM |
 | grok-4.6 | Maximum/Premium | $0.006 | Grok |
 | grok-4.5 | Standard/Economy | $0.006 | Grok |
@@ -174,7 +176,8 @@ the providers' official pricing pages: [Anthropic](https://platform.claude.com/d
 [MiniMax](https://platform.minimax.io/docs/guides/pricing-paygo). Gemini 3.7's
 listed rate is the promotion effective through December 31, 2026.
 
-`glm-5.3`, `qwen3.8-max`, and `kimi-code/k3-256k` are explicitly unpriced on
+`glm-5.3-flash`, legacy-pinned `glm-5.3`, `qwen3.8-max`, and
+`kimi-code/k3-256k` are explicitly unpriced on
 the transports used here. Their zero catalog entries prevent the estimator from
 inventing a dollar figure; session reports disclose their exclusion rather than
 presenting them as free.

package/docs/SETUP.md

@@ -396,6 +396,17 @@ INVOKING_AGENT=codex ./scripts/consult_all.sh "Question"
 
 Seven consultants can switch between CLI mode (using local CLI tools) and API mode (direct API calls): **Gemini, Codex, Claude, Mistral, Qwen3, Grok, and MiniMax**. Grok is CLI-first and runs in an isolated, tool-free strict sandbox. It falls back automatically only when Grok Build is missing, cannot launch, or lacks usable authentication and a key is available; request failures after launch are surfaced without an API fallback.
 
+Grok's normal OAuth mode is `GROK_OAUTH_MODE=shared`, which is safe for the
+parallel panel: each call gets a private HOME/workspace/prompt/output, while a
+runner-owned persistent `GROK_HOME` lets Grok Build coordinate token refresh
+with its own lock. A short ai-consultants lock atomically seeds and publishes
+valid generations; a concurrent external `grok login` wins the CAS and makes
+the affected run fail without an API fallback. Use
+`GROK_OAUTH_MODE=serialized` only as a diagnostic fallback. Forced API mode is
+stateless and does not read or create the shared OAuth generation. A newly
+created shared generation runs one locked inventory bootstrap to let Grok 1.0.4
+initialize its home metadata; the lock is released before inference.
+
 ### Why Use API Mode?
 
 - **No CLI installation required**: Use API keys without installing CLI tools

package/docs/cost_rates.json

@@ -1,5 +1,5 @@
 {
-  "version": "4.0.1",
+  "version": "4.0.3",
   "updated": "2026-08-23",
   "models": {
     "_comment_units": "All rates are USD per 1,000 tokens (per-1K). lib/costs.sh computes cost as (token_count / 1000) * rate, so every value here MUST be a per-1K rate = (provider $/1M price) / 1000. Do NOT paste $/MTok figures directly.",
@@ -402,6 +402,10 @@
       "input": 0,
       "output": 0
     },
+    "glm-5.3-flash": {
+      "input": 0,
+      "output": 0
+    },
     "kimi-code/k3-256k": {
       "input": 0,
       "output": 0
@@ -411,6 +415,7 @@
     "qwen3.8-max",
     "qwen3.8-max-preview",
     "glm-5.3",
+    "glm-5.3-flash",
     "kimi-code/k3-256k"
   ],
   "consultant_fallbacks": {
@@ -420,7 +425,7 @@
     "codex": "gpt-5.6-sol",
     "mistral": "mistral-large-3",
     "deepseek": "deepseek-v4-pro",
-    "glm": "glm-5.3",
+    "glm": "glm-5.3-flash",
     "grok": "grok-4.6",
     "qwen3": "qwen3.7-max",
     "kimi": "kimi-code/k3",
@@ -434,7 +439,7 @@
       "codex": "gpt-5.6-sol",
       "mistral": "mistral-medium-3.5",
       "deepseek": "deepseek-v4-pro",
-      "glm": "glm-5.3",
+      "glm": "glm-5.3-flash",
       "grok": "grok-4.6",
       "qwen3": "qwen3.8-max",
       "kimi": "kimi-code/k3-256k",
@@ -446,7 +451,7 @@
       "codex": "gpt-5.6-sol",
       "mistral": "mistral-medium-3.5",
       "deepseek": "deepseek-v4-pro",
-      "glm": "glm-5.3",
+      "glm": "glm-5.3-flash",
       "grok": "grok-4.6",
       "qwen3": "qwen3.7-max",
       "kimi": "kimi-code/k3",
@@ -458,7 +463,7 @@
       "codex": "gpt-5.6-terra",
       "mistral": "mistral-medium-3.5",
       "deepseek": "deepseek-v4-flash",
-      "glm": "glm-5.3",
+      "glm": "glm-5.3-flash",
       "grok": "grok-4.5",
       "qwen3": "qwen3.6-35b-a3b",
       "kimi": "kimi-code/k3",

package/docs/releases/v4.0.2.md

@@ -0,0 +1,98 @@
+# Release v4.0.2
+
+**Date:** 2026-08-28
+
+**Type:** Patch — concurrency-safe Grok OAuth
+
+**Previous:** v4.0.1
+
+## Highlights
+
+- Parallel Grok Build consultations can now share rotating OAuth safely without sharing agent state.
+- Refreshed credentials survive temporary HOME cleanup and are published atomically to the ambient session.
+- A concurrent external `grok login` always wins; the affected consultation fails without overwriting it or creating an API charge.
+- A real two-process `grok-4.6` smoke returned two structured `PONG` responses with exit 0.
+
+## What's New
+
+### Shared OAuth by default
+
+`GROK_OAUTH_MODE=shared` is now the normal Grok CLI path. The runner maintains one persistent credential generation under:
+
+```text
+$XDG_DATA_HOME/ai-consultants/grok-shared-oauth
+```
+
+Only `GROK_HOME` is shared. Each consultation still receives an independent HOME, workspace, private prompt, output, capability-probe environment, strict sandbox, empty tool surface, model, and effort.
+
+Grok Build's own auth lock coordinates token refresh between concurrent processes. The ai-consultants lock is used only for bounded seed, first-use bootstrap, authenticated inventory, and publication — never for inference.
+
+### Atomic refresh publication and CAS
+
+The shared marker binds the active generation to the SHA-256 of the credential last synchronized with ambient `~/.grok/auth.json`.
+
+After a run:
+
+- unchanged valid credentials require no write;
+- a valid rotation is copied into a mode-600 candidate and atomically renamed;
+- malformed rotation fails closed;
+- a changed ambient digest means an external login won, so the run fails temporarily and publishes no successful answer;
+- a publication failure never triggers the separately billed xAI API fallback.
+
+`GROK_OAUTH_MODE=serialized` remains available for diagnostic comparison and holds the runner lock across the complete CLI call.
+
+### Provider-owned configuration
+
+Grok 1.0.4 lazily creates provider-required metadata, caches, DBs, and `config.toml`. ai-consultants therefore preserves that file and stores its own atomic contract in `.ai-consultants-policy.json`. The effective agent restrictions remain pinned through the invocation arguments.
+
+Generations produced by the earlier development runner are recognized from their exact legacy config signature and superseded without deletion or mutation.
+
+### Sanitized failures
+
+Grok requests use redacted retry logging. Token-shaped provider stderr remains only in ephemeral private storage and never reaches argv, stdout, logs, error envelopes, or consultation artifacts.
+
+## Breaking Changes
+
+None. `shared` replaces the unsafe disposable-copy behavior for CLI OAuth, but model, effort, permissions, sandbox, response schema, API transport, and fallback classification remain unchanged.
+
+Users who need full-run serialization for diagnosis can set:
+
+```bash
+GROK_OAUTH_MODE=serialized
+```
+
+## Upgrade Guide
+
+```bash
+# npx, after npm publishes 4.0.2
+npx ai-consultants@4.0.2 version
+
+# active git-clone installation
+cd ~/.claude/skills/ai-consultants && git pull
+```
+
+No login or configuration migration is required. The first shared run seeds from the existing ambient Grok credential. No API key is copied or created.
+
+## Verification
+
+- Grok adapter: 88/88 checks
+- Configurator: 91/91 checks
+- Core gate: 452/452 checks
+- Full local gate: 23/23 suites
+- ShellCheck, JSON validation, npm tarball, and `git diff --check`: passed
+- PR #14 CI on exact head `883fad7`: passed
+- Real concurrent smoke: two exit-0 structured `PONG` responses, CLI transport, `grok-4.6`, `xhigh`, no workspace/runtime/lock residue, no API fallback, valid and hash-aligned ambient/generation/marker state
+
+## Provider-Specific Residuals
+
+- Real token rotation during concurrent inference was not forced; rotation/reuse is verified synthetically.
+- A real concurrent external `grok login` was not induced; external-login precedence is verified synthetically.
+- Superseded generations are retained to avoid deleting state still referenced by an active process.
+
+## Commits
+
+- `6c9e4b2` Merge pull request #14
+
+## Contributors
+
+- @matteoscurati

package/docs/releases/v4.0.3.md

@@ -0,0 +1,62 @@
+# Release v4.0.3
+
+**Date:** 2026-08-29
+**Type:** Patch — upgrade the GLM consultant to GLM-5.3-Flash
+**Previous:** v4.0.2
+
+## Highlights
+
+- GLM uses `glm-5.3-flash` in maximum, premium, and standard tiers.
+- `max_quality` continues to request GLM's `max` reasoning effort; the economy tier remains `glm-4-flash`.
+- Existing managed GLM defaults migrate forward while explicit model pins survive.
+- A real single-attempt Z.AI smoke confirmed the exact model and API transport.
+
+## What's New
+
+### GLM-5.3-Flash default
+
+New installations and runs without a GLM model override now request `glm-5.3-flash`. The same exact identifier is selected by the maximum, premium, and standard model tiers. This release does not change the Z.AI OpenAI-compatible endpoint, response budget, panel composition, routing affinities, or GLM's Methodologist persona.
+
+`max_quality` still sends `reasoning_effort=max` and a 16,384-token response budget. The economy preset remains on `glm-4-flash`.
+
+### Managed migration and cost disclosure
+
+When `ai-consultants configure` rewrites a configuration, the exact historical generated values `glm-5.2` and `glm-5.3` advance to `glm-5.3-flash`. Values marked `# ai-consultants:pin` remain untouched.
+
+GLM-5.3-Flash is explicitly listed as unpriced on the Coding Plan transport. The catalog also keeps legacy `glm-5.3` pins unpriced, preventing the estimator from presenting a fabricated fallback dollar rate.
+
+### Exact-transport verification
+
+One authorized live smoke used the shipped `query_glm.sh` adapter, the Z.AI OpenAI-compatible endpoint, `glm-5.3-flash`, max effort, a single attempt, and no retry. It completed with HTTP 200 and exit 0 in about 12 seconds; the response reported the exact requested model and measured 37 input plus 308 output tokens.
+
+The model returned near-structured text rather than the requested JSON. The adapter handled that response as designed: it retained the `PONG` content, labeled the answer as an unstructured-provider fallback, reduced confidence to 5, and disclosed the format caveat. This smoke therefore attests transport, authentication, model identity, max-effort acceptance, token accounting, and fallback behavior; it does not claim strict JSON adherence from every GLM-5.3-Flash response.
+
+## Breaking Changes
+
+None — this release is backwards-compatible. Explicit GLM model pins are preserved.
+
+## Upgrade Guide
+
+```bash
+# npx (no install)
+npx ai-consultants "question"
+
+# git clone install
+cd ~/.claude/skills/ai-consultants && git pull
+
+# curl | bash install
+curl -fsSL https://raw.githubusercontent.com/matteoscurati/ai-consultants/main/scripts/install.sh | bash
+```
+
+New installations use GLM-5.3-Flash automatically. Existing installations with an unpinned managed GLM value should run:
+
+```bash
+ai-consultants configure --force
+```
+
+Alternatively, set `GLM_MODEL=glm-5.3-flash` explicitly. A value marked `# ai-consultants:pin` will not be migrated.
+
+## Commits
+
+- `fc6aa85` feat: upgrade GLM default to 5.3 Flash
+- `b2690c3` Merge pull request #15 from matteoscurati/agent/glm-5.3-flash

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-consultants",
-  "version": "4.0.1",
+  "version": "4.0.3",
   "description": "A cross-vendor panel of up to 10 AI models that returns the coverage union of their distinct perspectives",
   "bin": {
     "ai-consultants": "bin/ai-consultants"

package/references/configuration.md

@@ -252,13 +252,14 @@ KIMI_MODEL=kimi-code/k3
 QWEN3_MODEL=qwen3.7-max
 QWEN3_MAX_QUALITY_TIMEOUT=600
 QWEN3_API_MAX_TOKENS=16384
-GLM_MODEL=glm-5.3
+GLM_MODEL=glm-5.3-flash
 GLM_API_MAX_TOKENS=16384
 GLM_REASONING_EFFORT=       # max_quality sets max
 GROK_MODEL=grok-4.6
 GROK_MAX_TURNS=4
 GROK_API_MAX_TOKENS=4096
 GROK_REASONING_EFFORT=      # max_quality sets xhigh (Grok Build maximum)
+GROK_OAUTH_MODE=shared      # shared (parallel default) | serialized (diagnostic)
 DEEPSEEK_MODEL=deepseek-v4-pro
 DEEPSEEK_API_MAX_TOKENS=16384
 DEEPSEEK_MAX_QUALITY_TIMEOUT=600
@@ -285,6 +286,28 @@ Set `GROK_USE_API=true` only to force the API path. The adapter accepts any CLI
 version that exposes the complete required headless interface and requested
 model; `metadata.cli_version` is provenance, not a version pin.
 
+`GROK_OAUTH_MODE=shared` is the normal CLI path because ai-consultants launches
+consultants in parallel. Each Grok invocation keeps a private HOME, workspace,
+prompt, output, permissions, and tool-free agent policy. Concurrent Grok
+processes share only a runner-owned persistent `GROK_HOME` generation under
+`$XDG_DATA_HOME/ai-consultants/grok-shared-oauth` (or the standard XDG data
+fallback). Grok Build's own auth lock coordinates refreshes inside that shared
+generation; a short ai-consultants lock protects seeding, adoption, atomic
+runner-policy reconciliation, and atomic publication back to the ambient
+`~/.grok/auth.json`. An external `grok login` always wins the digest-based CAS:
+the affected consultation fails temporarily and never overwrites the new login
+or falls back to a separately billed API request. Malformed credentials and
+unpersisted refreshes fail closed. Use `GROK_OAUTH_MODE=serialized` only to hold
+the adapter lock across the complete CLI run for diagnostic comparison. Forced
+API mode is stateless and never creates or reads the shared OAuth generation.
+Each new generation also performs exactly one model-inventory bootstrap under
+the short lock because Grok 1.0.4 lazily initializes home metadata; all later
+inventory probes and every inference remain concurrent.
+Grok 1.0.4 retains ownership of its generated `config.toml`; ai-consultants does
+not replace provider-required sections. Instead it atomically reconciles a
+separate `.ai-consultants-policy.json`, while the actual enforced surface stays
+pinned by the per-invocation CLI arguments.
+
 `max_quality` sets `GROK_REASONING_EFFORT=xhigh` (the highest value accepted by
 Grok Build),
 `GLM_REASONING_EFFORT=max`, and `DEEPSEEK_REASONING_EFFORT=max`. Grok passes

package/scripts/config.sh

@@ -186,7 +186,7 @@ QWEN3_FORMAT="${QWEN3_FORMAT:-qwen}"
 # GLM CONFIGURATION - The Code Specialist (API-based)
 # =============================================================================
 
-GLM_MODEL="${GLM_MODEL:-glm-5.3}"
+GLM_MODEL="${GLM_MODEL:-glm-5.3-flash}"
 GLM_TIMEOUT_SECONDS="${GLM_TIMEOUT:-180}"
 GLM_API_MAX_TOKENS="${GLM_API_MAX_TOKENS:-16384}"
 GLM_API_URL="${GLM_API_URL:-https://api.z.ai/api/coding/paas/v4/chat/completions}"
@@ -202,6 +202,7 @@ GROK_MODEL="${GROK_MODEL:-grok-4.6}"
 GROK_TIMEOUT_SECONDS="${GROK_TIMEOUT:-180}"
 GROK_API_MAX_TOKENS="${GROK_API_MAX_TOKENS:-4096}"
 GROK_MAX_TURNS="${GROK_MAX_TURNS:-4}"
+GROK_OAUTH_MODE="${GROK_OAUTH_MODE:-shared}"
 GROK_API_URL="${GROK_API_URL:-https://api.x.ai/v1/chat/completions}"
 GROK_FORMAT="${GROK_FORMAT:-openai}"
 # Grok Build's official credential name is XAI_API_KEY. Keep the historical
@@ -221,7 +222,7 @@ if [[ -z "${GROK_USE_API+x}" ]]; then
         GROK_USE_API=false
     fi
 fi
-export GROK_USE_API
+export GROK_USE_API GROK_OAUTH_MODE
 
 # =============================================================================
 # DEEPSEEK CONFIGURATION - The Methodologist (API-based)
@@ -622,7 +623,7 @@ get_model_for_tier() {
                 codex)    echo "gpt-5.6-sol" ;;
                 mistral)  [[ "$transport" == "api" ]] && echo "mistral-large-3" || echo "mistral-medium-3.5" ;;
                 deepseek) echo "deepseek-v4-pro" ;;
-                glm)      echo "glm-5.3" ;;
+                glm)      echo "glm-5.3-flash" ;;
                 grok)     echo "grok-4.6" ;;
                 qwen3)    echo "qwen3.8-max" ;;
                 kimi)     echo "kimi-code/k3-256k" ;;
@@ -637,7 +638,7 @@ get_model_for_tier() {
                 codex)    echo "gpt-5.6-sol" ;;
                 mistral)  [[ "$transport" == "api" ]] && echo "mistral-large-3" || echo "mistral-medium-3.5" ;;
                 deepseek) echo "deepseek-v4-pro" ;;
-                glm)      echo "glm-5.3" ;;
+                glm)      echo "glm-5.3-flash" ;;
                 grok)     echo "grok-4.6" ;;
                 qwen3)    echo "qwen3.7-max" ;;
                 kimi)     echo "kimi-code/k3" ;;
@@ -652,7 +653,7 @@ get_model_for_tier() {
                 codex)    echo "gpt-5.6-terra" ;;
                 mistral)  [[ "$transport" == "api" ]] && echo "mistral-large-3" || echo "mistral-medium-3.5" ;;
                 deepseek) echo "deepseek-v4-flash" ;;
-                glm)      echo "glm-5.3" ;;  # Same as premium (no mid-tier GLM)
+                glm)      echo "glm-5.3-flash" ;;  # Same as premium (no mid-tier GLM)
                 grok)     echo "grok-4.5" ;;
                 qwen3)    echo "qwen3.6-35b-a3b" ;;  # Open-weight MoE (35B total, 3B active)
                 kimi)     echo "kimi-code/k3" ;;
@@ -1010,4 +1011,4 @@ EOF
 # VERSION
 # =============================================================================
 
-AI_CONSULTANTS_VERSION="4.0.1"
+AI_CONSULTANTS_VERSION="4.0.3"

package/scripts/configure.sh

@@ -307,11 +307,14 @@ if [[ "$(read_value "$WORK_FILE" GEMINI_MODEL 2>/dev/null || true)" == "Gemini 3
     echo "Migrated managed Gemini CLI default: Gemini 3.1 Pro (High) -> Gemini 3.7 Flash (High)" >&2
 fi
 
-if [[ "$(read_value "$WORK_FILE" GLM_MODEL 2>/dev/null || true)" == "glm-5.2" ]] \
-    && ! has_value_marker "$WORK_FILE" GLM_MODEL "$PIN_MARKER"; then
-    set_value GLM_MODEL "glm-5.3 $DEFAULT_MARKER"
-    echo "Migrated managed GLM default: glm-5.2 -> glm-5.3" >&2
-fi
+case "$(read_value "$WORK_FILE" GLM_MODEL 2>/dev/null || true)" in
+    glm-5.2|glm-5.3)
+        if ! has_value_marker "$WORK_FILE" GLM_MODEL "$PIN_MARKER"; then
+            set_value GLM_MODEL "glm-5.3-flash $DEFAULT_MARKER"
+            echo "Migrated managed GLM default to glm-5.3-flash" >&2
+        fi
+        ;;
+esac
 
 if [[ "$(read_value "$WORK_FILE" GROK_MODEL 2>/dev/null || true)" == "grok-4.5" ]] \
     && ! has_value_marker "$WORK_FILE" GROK_MODEL "$PIN_MARKER"; then

package/scripts/lib/common.sh

@@ -221,7 +221,9 @@ run_query() {
 
         # Error handling
         local error_msg=""
-        [[ -f "$error_file" ]] && error_msg=$(head -5 "$error_file" 2>/dev/null)
+        if [[ "${RUN_QUERY_REDACT_ERRORS:-false}" != "true" && -f "$error_file" ]]; then
+            error_msg=$(head -5 "$error_file" 2>/dev/null)
+        fi
 
         # First line of the CLI's own stderr — the actual reason (auth error,
         # command not found, etc.). Kept for the final failure message so it