0xray 3.5.4 → 4.0.0

package/AGENTS-consumer.md

@@ -4,7 +4,15 @@
 
 Quick reference for the 0xRay AI orchestration framework (shipped to your project on `npm install 0xray`).
 
-**v3.5.3** — 7 MCP servers · 45 skills · 68 codex terms · 4 platform bridges
+**v4.0.0** — a suit that survives the context window.
+
+**v4.0.0** — same **three-subsystem OS** as v2 (Inference · External Governance · Autonomous Engine). Constitution always on. Temperament scales ceremony. Four platform bridges. Repertoire auto-enables when the module is present; otherwise `memory_routing` stays off. Trim fat, do not gut the design.
+
+**Governance always on** (Codex PreToolUse, no new surface, no `eval`). **Engine ceremony** (analyze-complexity before spawn, confer) is **guided** by default so free-model OpenCode/Hermes stay in check. Frontier hosts (Grok 4.6 class) with `suit_temperament.profile: auto` **warn** on spawn-without-plan instead of deny — the engine is still there. Existing `.xray/features.json` without `suit_temperament` stays **guided**. Docs: v3 from v2 · Suit temperament.
+
+## Hot-swap
+
+Before other work, Read `.xray/state/STATION.md`. Compaction and host change are the same cut. Continue the card. Do not restart the job. Grok does not inject the card (OpenCode does) — this Read is the Grok contract. Do not thicken the Grok exo.
 
 ## Postinstall (automatic)
 
@@ -76,14 +84,16 @@ Configure in `.xray/features.json`:
 "memory_routing": {
   "enabled": true,
   "provider": "repertoire",
-  "module_path": "../repertoire/dist/provider/memory-routing-provider.js",
+  "module_path": "node_modules/@0xray/repertoire/dist/provider/memory-routing-provider.js",
   "config": {
-    "signalsPath": "../repertoire/data/curated_signals.json"
+    "signalsPath": "node_modules/@0xray/repertoire/data/curated_signals.json",
+    "statePath": ".xray/state/repertoire/inference-state.json",
+    "feedbackDir": ".xray/state/repertoire/feedback"
   }
 }
 ```
 
-Without Repertoire: `{ "enabled": false, "provider": "null" }`.
+Opt out: `{ "enabled": false, "provider": "repertoire" }`.
 
 **External MCP** (Hermes/Grok): add Repertoire alongside 0xRay servers:
 
@@ -104,7 +114,7 @@ The Universal Development Codex (`.xray/codex.json`, **69 terms**) is enforced b
 |------|------|-------------|
 | 11 | No `any`, `@ts-ignore`, `@ts-expect-error` | PreToolUse deny on edits |
 | 29 | Security by design — no `eval()` | PreToolUse deny |
-| 59 | Complex work → orchestrator intake | PreToolUse deny `spawn_subagent` without plan |
+| 59 | Complex work → orchestrator intake | **guided/strict:** PreToolUse **deny** spawn without plan. **frontier:** **warn** (allow) — engine still available; leftover-plan todo mismatch also warns. Codex 11/29/69 still **deny**. |
 | 67–68 | Best subagents + lead dev ownership | `analyze-complexity` + orchestrator skill |
 | **69** | **No new MCP/skill/handler surface** | PreToolUse deny new `*.server.ts`, `SKILL.md` |
 
@@ -123,7 +133,7 @@ When the suit is worn, **lead dev mode** is ON via existing config + hooks:
 | Test triage | PreToolUse hook | Per-suite hint on full `npm test` |
 | Playbook | **`orchestrator`** skill | Codex 59, 67–69 rules |
 
-**First substantive task:** `analyze-complexity` with `tasks` array — **required before** `spawn_subagent`. You are the **lead developer**. Users speak in goals, not keywords.
+**First substantive task (guided/strict):** `analyze-complexity` with `tasks` array — **required before** `spawn_subagent`. **Frontier:** intake is optional; spawn without a plan **warns**. You are the **lead developer**. Users speak in goals, not keywords.
 
 | # | Rule |
 |---|------|

package/AGENTS.md

@@ -1,15 +1,23 @@
 # xray Agents
 
-Quick reference for the xray AI orchestration framework (**v3.5.3**).
+Quick reference for the xray AI orchestration framework (**v4.0.0**).
 
-**42 agents** · **45 skills** · **7 consumer MCP servers** · **68 codex terms** · **3,226 tests**
+**v4.0.0** — a suit that survives the context window.
+
+**Exo, not catalog.** Three-subsystem OS (Inference · External Governance · Autonomous Engine) · constitution always on · temperament by host · 4 floors · Repertoire when it resolves.
+
+**Temperament:** ceremony (analyze-complexity before spawn, confer) is **guided** by default so free-model OpenCode/Hermes stay in check. Frontier hosts (Grok 4.6 class) with `suit_temperament.profile: auto` **warn** on spawn-without-plan; Codex 11/29/69 still **deny**. Missing `suit_temperament` stays guided. See `docs-site/docs/architecture/v3-from-v2.md`.
+
+## Hot-swap
+
+Before other work, Read `.xray/state/STATION.md`. Compaction and host change are the same cut. Continue the card. Do not restart the job. Grok does not inject the card (OpenCode does) — this Read is the Grok contract. Do not thicken the Grok exo.
 
 ## What is xray?
 
 xray provides intelligent multi-agent orchestration with automatic delegation and Codex compliance validation under the pure v2 three-subsystem model:
 
 - **Inference** — proposals, reflection, memory routing (optional)
-- **External Governance** — Dynamo Solar SSOT, 68-term Codex
+- **External Governance** — Dynamo Solar SSOT, 69-term Codex
 - **Autonomous Engine** — thinDispatch 7-flow, AsideContext, confidence gate
 
 Agents are declared in `src/opencode/agents/*.yml` — the YML SSOT. Skills live in `src/skills/*/SKILL.md` — see [SKILLS.md](SKILLS.md). Consumer projects receive slimmed copies via postinstall (`AGENTS-consumer.md` → `AGENTS.md`, skills synced to platform dirs).
@@ -77,6 +85,16 @@ Governance deliberation: **code-review**, **security-audit**, **researcher** wit
 - Observation extractors: governance, orchestration, complexity
 - Repertoire context flows via `ExecutionPlan.memoryContext`
 
+## User asides (parallel work tracks)
+
+**Not** AsideContext — user-declared parallel tracks (worktree-like):
+
+- SSOT: `.xray/state/asides/{id}.json` + `_active.json`; todos use `{asideId}.a.*` (namespaced)
+- Intake: `analyze-complexity` with `userAsideId` (+ optional `worktree`, `branch`)
+- Activate / resume main: `orchestrate-task` with `userAsideId` or `clearActiveAside: true`
+- Spawn gate routes to active aside plan; session-boot exposes `activeAside`
+- Config: `features.json` → `multi_agent_orchestration.user_asides.enabled`
+
 ## Confidence gate (v3.3.1)
 
 `ExecutionPlanner.calculateTaskComplexity()` + `getTaskConfidence()` — complexity boost and ontological-trap routing.

package/README.md

@@ -1,22 +1,24 @@
-# xray — MCP-Centric AI Governance OS
+# xray — exo for coding agents
 
-**v3.5.3** — 42 agents · 45 skills · 7 MCP servers · 68 codex terms · 3,226 tests
+**v4.0.0** — a suit that survives the context window. Constitution always on. Temperament scales ceremony. Four host floors. Repertoire is the muscle for handoffs.
+
+Not a catalog of agents. The product is the **skeleton you wear**.
 
 ## Your AI Power Suit
 
 **Your AI Power ⚡️ Suit 🦾** — an exoskeleton that blocks AI slop from becoming real damage.
 
-- learns with **Repertoire** 🧠
-- has a **Solar Ban Hammer** ☀️🔨 (Dynamo governance)
-- can also **Confer** 🗣️ (3-agent internal deliberation)
-
-The core AI Governance OS stack is **complete and open source**. A commercial version is next — powers up the suit further.
+- **Bone** — three-subsystem OS: Inference · External Governance (Dynamo / Codex) · Autonomous Engine (thinDispatch)
+- **Always on** — Codex 11 / 29 / 69, destructive shell, no new MCP/skill/handler surface
+- **Temperament** — `frontier` | `guided` | `strict` | `auto` — how loud the engine is, not whether governance exists
+- **Wear** — Grok, OpenCode, Hermes, OpenClaw. One SSOT gate
+- **Muscle** — Repertoire: auto-enables when the module resolves; session-start writes a one-line resume so the next station is not born amnesiac
 
 ```bash
-npm install 0xray@latest
+npm install 0xray
 ```
 
-xray is the pure v2 three-subsystem AI orchestration framework — **MCP-centric**, governed by Dynamo, and autonomous via thinDispatch. Consumer `npm install 0xray` auto-wires all four platform bridges and seven MCP servers with zero manual config.
+Consumer `npm install 0xray` auto-wires four platform bridges and seven MCP servers. Repertoire **0.2** is the factory memory organ (vendored): seed primitives, in-process routing, extra host MCP `repertoire`. Not an eighth `xray-*` server. Opt out with `"enabled": false, "provider": "repertoire"`.
 
 ## Quick Start
 
@@ -46,6 +48,7 @@ Docs: [guides/autonomy-command](docs-site/docs/guides/autonomy-command.md) · Sk
 
 | Version | Highlights |
 |---------|------------|
+| **4.0.0** | Exo + temperament. Constitution always on; ceremony by host. Repertoire enable-when-resolves + session resume. Survive the context window — still in development, not on npm yet. [vision](docs-site/docs/architecture/v4-vision.md) · [now](docs-site/docs/architecture/v4-now.md) |
 | **3.4.1** | Unified `install-bridges.cjs` on postinstall — OpenCode, Grok, Hermes, OpenClaw in one pass. All 7 MCPs servers via `npx -y 0xray mcp <cmd>` (no `dist/` paths). Canonical `release.mjs` pipeline. |
 | **3.3.1** | Orchestrator confidence gate wired into execution planning. |
 | **3.3.0** | Pluggable **Memory Routing** (`features.json` → `memory_routing`). Repertoire is the default provider in the framework repo. |
@@ -165,7 +168,7 @@ Pluggable `memory_routing` block in `features.json` (validated by `features.sche
 
 **External hosts** (Hermes/Grok): add `repertoire-mcp` to `.mcp.json` — see Repertoire docs.
 
-Consumers without Repertoire: `"memory_routing": { "enabled": false, "provider": "null" }`.
+Opt out: `"memory_routing": { "enabled": false, "provider": "repertoire" }`.
 
 Docs: [memory routing](docs-site/docs/guides/memory-routing.md) · [Repertoire](docs-site/docs/guides/repertoire.md) · [features.json](docs-site/docs/guides/features-json.md) · [all features since 3.1](docs-site/docs/guides/features-since-3.1.md)
 

package/SKILLS.md

@@ -1,6 +1,6 @@
 # 0xray Skills
 
-**v3.5.3** — **45 skills** · invoked via `xray-skills` MCP · synced on postinstall
+**v4.0.0** — **45 skills** · invoked via `xray-skills` MCP · synced on postinstall
 
 xray ships 45 knowledge skills (`src/skills/<name>/SKILL.md`) across the full software lifecycle. Lead-dev operating model lives in **`orchestrator`** skill + `multi_agent_orchestration.lead_dev_mode` in features.json. Skills load on demand via the **xray-skills** MCP server (`npx -y 0xray mcp skills`), agent `@mentions`, or platform skill directories after postinstall.
 

package/dist/AGENTS.md

@@ -1,15 +1,23 @@
 # xray Agents
 
-Quick reference for the xray AI orchestration framework (**v3.5.3**).
+Quick reference for the xray AI orchestration framework (**v4.0.0**).
 
-**42 agents** · **45 skills** · **7 consumer MCP servers** · **68 codex terms** · **3,226 tests**
+**v4.0.0** — a suit that survives the context window.
+
+**Exo, not catalog.** Three-subsystem OS (Inference · External Governance · Autonomous Engine) · constitution always on · temperament by host · 4 floors · Repertoire when it resolves.
+
+**Temperament:** ceremony (analyze-complexity before spawn, confer) is **guided** by default so free-model OpenCode/Hermes stay in check. Frontier hosts (Grok 4.6 class) with `suit_temperament.profile: auto` **warn** on spawn-without-plan; Codex 11/29/69 still **deny**. Missing `suit_temperament` stays guided. See `docs-site/docs/architecture/v3-from-v2.md`.
+
+## Hot-swap
+
+Before other work, Read `.xray/state/STATION.md`. Compaction and host change are the same cut. Continue the card. Do not restart the job. Grok does not inject the card (OpenCode does) — this Read is the Grok contract. Do not thicken the Grok exo.
 
 ## What is xray?
 
 xray provides intelligent multi-agent orchestration with automatic delegation and Codex compliance validation under the pure v2 three-subsystem model:
 
 - **Inference** — proposals, reflection, memory routing (optional)
-- **External Governance** — Dynamo Solar SSOT, 68-term Codex
+- **External Governance** — Dynamo Solar SSOT, 69-term Codex
 - **Autonomous Engine** — thinDispatch 7-flow, AsideContext, confidence gate
 
 Agents are declared in `src/opencode/agents/*.yml` — the YML SSOT. Skills live in `src/skills/*/SKILL.md` — see [SKILLS.md](SKILLS.md). Consumer projects receive slimmed copies via postinstall (`AGENTS-consumer.md` → `AGENTS.md`, skills synced to platform dirs).
@@ -77,6 +85,16 @@ Governance deliberation: **code-review**, **security-audit**, **researcher** wit
 - Observation extractors: governance, orchestration, complexity
 - Repertoire context flows via `ExecutionPlan.memoryContext`
 
+## User asides (parallel work tracks)
+
+**Not** AsideContext — user-declared parallel tracks (worktree-like):
+
+- SSOT: `.xray/state/asides/{id}.json` + `_active.json`; todos use `{asideId}.a.*` (namespaced)
+- Intake: `analyze-complexity` with `userAsideId` (+ optional `worktree`, `branch`)
+- Activate / resume main: `orchestrate-task` with `userAsideId` or `clearActiveAside: true`
+- Spawn gate routes to active aside plan; session-boot exposes `activeAside`
+- Config: `features.json` → `multi_agent_orchestration.user_asides.enabled`
+
 ## Confidence gate (v3.3.1)
 
 `ExecutionPlanner.calculateTaskComplexity()` + `getTaskConfidence()` — complexity boost and ontological-trap routing.

package/dist/CHANGELOG.md

@@ -4,13 +4,111 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Conventional Commits](https://www.conventionalcommits.org/).
 
+## [4.0.0] - 2026-08-28
+
+First ship of the **temperament** line on the **v2 three-subsystem OS** (Inference · External Governance · Autonomous Engine). npm is still `3.5.5`; this line is **4.0.0** on `feat/v4-temperament` — **still in development, not published**.
+
+### ✨ Features
+- **Factory memory organ** — `@0xray/repertoire@0.2.0` vendored as a runtime dependency. Shipped `memory_routing` is **on** (8 factory primitives). Extra host MCP stays `repertoire`, not an 8th `xray-*` server. Compact ingest writes `.xray/state/repertoire/`. Explicit opt-out: `enabled: false` + `provider: "repertoire"`.
+- **Four-floor compact + memory** — station heat matches Repertoire on every host (not only Grok). Compact on Grok Pre/PostCompact, Hermes session-start, OpenCode plugin boot, OpenClaw PreToolUse keeps intent, Working signals, and `Repertoire: on`.
+- **Suit temperament** — `frontier` | `guided` | `strict` | `auto`. Constitution (Codex 11/29/69, destructive shell) always on. Ceremony (spawn-plan deny, confer, synthesis) scales by host.
+- **Grok 4.6 / Grok Build** — `auto` → frontier: spawn missing/stale/todo-mismatch **warns**; confer off unless opted in.
+- **OpenCode / Hermes** (including free models) — stay **guided**: full v2 engine ceremony.
+- **Upgrade safety** — consumers without `suit_temperament` stay **guided**; postinstall merge will not frontier existing Grok installs.
+- **Codex 69** — in `codex.json` (blocking). PreToolUse already denied new MCP/skill/handler files.
+- **OpenClaw host PreToolUse** — `evaluateOpenClawHostPreTool` SSOT + `~/.openclaw/hooks/xray-pre-tool.mjs` on `0xray openclaw install`. Platform still has no native abort API; this is the host adapter.
+- **Repertoire harness** — leftover `memory_routing` default-off auto-enables when sibling `@0xray/repertoire` or `node_modules/@0xray/repertoire` resolves (install **and** runtime provider load). Explicit `enabled: false` + `provider: repertoire` is opt-out. Consumers without the module stay off. Not an 8th 0xRay MCP.
+- **Repertoire working state** — station heat writes `.xray/state/repertoire-working.json` and a `Working:` line on the card. Compact may ingest non-Bedrock matches. Registry primitives are routing; the working file is session memory.
+- **Session resume** — Grok session-start writes `repertoireResume` (one line, signal count) into `session-boot.json` so the next station survives compaction.
+- **Hot-swap station card** — `.xray/state/STATION.md` is the projection of `session-boot.json` (intent, git, **live** plan, last host, Repertoire). PreCompact/PostCompact persist it. AGENTS.md tells the model to Read it (Grok does not inject). Host change sets `hotSwap`. Same-host compact **keeps** the last swap stamp. OpenClaw first PreToolUse heats the card. Intent strips `<user_query>` wrappers. Works without Repertoire.
+- **Framework dogfood wear** — `installAllBridges` on the 0xRay package still patches `<project>/.grok/hooks/0xray.json` and enable-when-resolves. Does not deploy consumer copies over the package.
+- **Repertoire MCP names** — unprefixed server tools; Grok TUI namespaces them as `repertoire__*`. Grok install resolves sibling *or* `node_modules/@0xray/repertoire`.
+- **Configurable archive marker** — `plan_archive_marker_hours` (default 24) in `features.json`
+- **grok_postprocessor_light** — PostToolUse write-tool marker + activity.log
+- **P2 session capture** — git post-commit → `captureSessionInference` when `inference_session_capture.min_commits` met
+- **Docs** — [4.0 vision](docs-site/docs/architecture/v4-vision.md), [v3 from v2](docs-site/docs/architecture/v3-from-v2.md), [temperament](docs-site/docs/guides/v3-temperament.md), [trim list](docs-site/docs/architecture/v3-museum.md).
+
+### 🔄 Changes
+- Product version **4.0.0** throughout (`package.json`, `features.json`, plugin, banners). 3.x entries below are historical (npm `3.0.0`–`3.5.5`).
+- `evaluatePreToolGate` is constitution-first (Grok, Hermes, OpenCode). Grok PreToolUse no longer re-runs 11/29/69/surface/destructive after the SSOT.
+- Codex 69 denies **new** surface only (existing MCP/skill files may be rewired). `paths[]` / `target_notebook` / `prompt` content are in the SSOT collector.
+- **strict** locks `no_new_surface` on (cannot opt out).
+- Dual-host session-boot: Hermes / OpenCode / OpenClaw write `host` + `suit_profile`. A Hermes host ignores leftover Grok boot (and vice versa). Kernel/`generic` uses last boot.
+- Frontier `auto_consult` follows confer opt-in (`confer.enabled` / `confer_on_synthesis`).
+- Session-boot refreshes leftover host/workspace from another machine.
+- Post-commit log maintenance prefers consumer `node_modules/0xray/dist` and ESM `pathToFileURL`.
+
+### 🐛 Fixes
+- OpenClaw plugin ships in the npm tarball; install resolves `dist` then `src`.
+- Bare `node scripts/node/release.mjs` no longer defaults to patch (that path applied npm `3.5.5` → `3.5.6` over the 4.0 plant). Bump type is required. CJS/MJS loadability syntax-checks version/release/reflection/prepare/setup/CI-report scripts instead of executing them.
+- Postinstall wears OpenClaw PreToolUse (stdin hook + plugin link), not only `0xray openclaw install`.
+- Ephemeral e2e consumers no longer overwrite machine Hermes/OpenClaw `xray-consumer-root.txt`.
+- Hermes postinstall copies hook runtimes; plugin enable prefers `xray-hermes`.
+- OpenClaw install writes `opencode-cli` backend when `opencode` is on PATH (CLI pickle, not Zen HTTP).
+- `0xray grok install` always runs `patchGrokHooks` (command strings, PreCompact/PostCompact) even when the plugin already exists. Stale `args[]` / smoke / blaze paths no longer leave compact cold. Patched JSON is also copied to `<project>/.grok/hooks/0xray.json` — the path Grok TUI actually loads.
+- `buildLeadDevPlan` / `buildSynthesisCheckpointPlan` resolve temperament from `projectRoot`, not `process.cwd()` session-boot. A Grok frontier checkout no longer empties guided confer todos in tmp asides and release-gate verify scripts.
+- Stop committing machine-local `.grok/` (absolute hook paths). Template stays `src/integrations/grok/plugin/0xray`.
+- Grok hook `command` strings quote `XRAY_AI_PATH` and the script path so installs with spaces still start PreToolUse (Grok fail-opens on crash).
+- OpenClaw `maybeHeatHostStation` skips only when both sides have a concrete matching session id. An install card with no session no longer swallows first PreToolUse.
+- `postinstall.cjs` always calls `installAllBridges` (framework dogfood wear is no longer gated out of the production caller).
+- `loadFeatures` forwards `grok_postprocessor_light`; PostToolUse can run the advertised write-tool marker path.
+- `0xray grok install` writes `[mcp_servers.repertoire]` to `<project>/.grok/config.toml` and resolves the sibling from the consumer root.
+- OpenClaw 2026.8.2 handshake uses client id `gateway-client` (not `openclaw-tui`). Control-UI id requires device identity; TUI id is rejected on loopback.
+- OpenClaw e2e falls back to `openclaw agent` when token-only WS lacks `operator.write`. Grok e2e asserts `evaluatePreToolGate` SSOT instead of inlined gate names.
+
+### ✅ Tests
+- `suit-temperament.test.ts`; leftover-plan warn; confer frontier opt-in; install-bridges guided pin; Codex 69 rewrite vs create; strict lock; `paths[]`; dual-host boot.
+- `spawn-plan-stale` after archival in `verify-delegation-gate-core` (6/6); OpenCode host in `release-gate`.
+- OpenClaw PreToolUse unit coverage; live temperament probe across Grok, Hermes, OpenCode, OpenClaw.
+- Wear-layout: plugin/hook resolve, ephemeral marker skip, opencode-cli backend write.
+- `grok-hooks-json`: stale `args[]` smoke hooks gain compact events; station compact keeps `hotSwap`.
+- `autonomy-kernel`: plan consults follow `projectRoot` profile; confer/synthesis verify scripts pass tmp, not cwd Grok boot.
+- Repertoire enable-when-resolves + session-start `repertoireResume` unit coverage.
+- Quoted Grok hook commands; postinstall dogfood wear; OpenClaw install-then-first-tool heat; `grok_postprocessor_light` loadFeatures; project Repertoire `config.toml`.
+- Four-floor compact survival attests Codex 11 / 69 / destructive shell plus temperament after the cut. `XRAY_LIVE_SUCCESSOR=1` asks each host CLI to quote the station card.
+
+---
+
+## [3.5.5] - 2026-06-20
+
+### ✨ Features
+- **User-asides P0** — PreToolUse **denies** aside spawns without `cwd` / `working_directory` matching `aside.worktree`; `aside-worktree.ts` SSOT; synthesis-gate parity
+- **Worktree auto-provision** — `analyze-complexity` runs guarded `git worktree add` when `user_asides.auto_provision_worktree: true`
+- **Pipeline P3** — `verify:pipeline-facets` script; consumer `confirm:suit:full --pipeline-audit` wiring
+
+### 🐛 Fixes
+- **Worktree cwd enforce** — `auto_chain_delegations: false` no longer bypasses aside worktree cwd check; synthesis-gate parity retained
+- **`isAutoProvisionWorktreeEnabled`** — requires `user_asides.enabled` before honoring `auto_provision_worktree`
+- **`provisionGitWorktree`** — macOS `/var` vs `/private/var` path resolution for idempotent worktree detection
+
+### 🔄 Changes
+- **`verify:pipeline-facets --package-only`** — release-gate probes shipped artifacts without consumer feature matrix
+- **`release-gate.mjs`** — wires pipeline facet verify (package-only mode)
+
+### ✅ Tests
+- `aside-worktree.test.ts`; delegation-gate cwd deny/allow; auto_chain-off + synthesis worktree deny; `provisionGitWorktree` guards; `verify-user-aside-core` 10/10
+
+---
+
 ## [3.5.4] - 2026-06-19
 
 ### ✨ Features
 - **Synthesis consult receipt gate** — s.1–s.3 todos require `.xray/state/synthesis-consult-{id}.json` before completion; `evaluatePostToolSpawn` records receipts from subagent output; Grok post-tool hook wired
+- **User asides** — parallel work tracks (worktree-like) via `.xray/state/asides/{id}.json`; namespaced `{asideId}.a.*` todo routing; `analyze-complexity` + `orchestrate-task` rewire (no new MCP); session-boot hints; synthesis/pending bypass for aside spawns; `verify:user-aside` 9/9; docs `guides/user-asides.md`
+- **Confer quorum** — `confer.ts` SSOT; auto-runs at synthesis `analyze-complexity` (researcher → architect-tools → code-review); `verify:confer` 7/7; config `confer_on_synthesis` (default true); session-boot `conferPending` hint; emoji agent panel in quorum report
+- **Pipeline integration (P0–P2)** — enforcement-gate `{ proposals: [...] }` shape; `applyDecisionMatrix` + `calculateMetamorphosisScore` in `GovernanceService`; Grok `pipeline-hook-runtime.mjs` (reflection stubs, routing outcomes, reporting, inference-improvement light); `SelfProposalEngine` wired in PostProcessor; aside worktree cwd warn in delegation-gate; `autonomous_reporting` config key fix in SessionSummaryProcessor
+- **Stale lead-dev plan** — archive after `plan_stale_hours`; `spawn-plan-stale` persists after archival via recent archive marker; synthesis realignment (s.1–s.3) exempt until consult todos complete
+- **Hermes session-start parity** — `bridge.mjs` `session-start` command archives stale plans; lazy `plan-hook-runtime.mjs`; `verify-hermes-session-start.mjs` (4/4)
+- **Pre-commit / pre-push codex** — diff-hunk and commit-range scope verifiers (`verify:pre-commit-diff`, `verify:pre-push-diff`)
+- **Reflection noise** — minimal defaults 50 commits / `autoGenerate: false`; shell fallbacks synced
+- **Multi-repo boot** — `sibling_repos` in `session-boot.json`
+
+### 🔄 Changes
+- **autonomy-command.md** — SSOT is `analyze-complexity` + `multi_agent_orchestration`; removed stale `autonomy-intake` / `autonomy_kernel` references
 
 ### ✅ Tests
 - `synthesis-consult-receipt.test.ts`; E2E/outcome verifiers write fixture receipts
+- `governance-service-metamorphosis.test.ts`; `confer.test.ts` emoji panel
 
 ---
 

package/dist/README.md

@@ -1,22 +1,24 @@
-# xray — MCP-Centric AI Governance OS
+# xray — exo for coding agents
 
-**v3.5.3** — 42 agents · 45 skills · 7 MCP servers · 68 codex terms · 3,226 tests
+**v4.0.0** — a suit that survives the context window. Constitution always on. Temperament scales ceremony. Four host floors. Repertoire is the muscle for handoffs.
+
+Not a catalog of agents. The product is the **skeleton you wear**.
 
 ## Your AI Power Suit
 
 **Your AI Power ⚡️ Suit 🦾** — an exoskeleton that blocks AI slop from becoming real damage.
 
-- learns with **Repertoire** 🧠
-- has a **Solar Ban Hammer** ☀️🔨 (Dynamo governance)
-- can also **Confer** 🗣️ (3-agent internal deliberation)
-
-The core AI Governance OS stack is **complete and open source**. A commercial version is next — powers up the suit further.
+- **Bone** — three-subsystem OS: Inference · External Governance (Dynamo / Codex) · Autonomous Engine (thinDispatch)
+- **Always on** — Codex 11 / 29 / 69, destructive shell, no new MCP/skill/handler surface
+- **Temperament** — `frontier` | `guided` | `strict` | `auto` — how loud the engine is, not whether governance exists
+- **Wear** — Grok, OpenCode, Hermes, OpenClaw. One SSOT gate
+- **Muscle** — Repertoire: auto-enables when the module resolves; session-start writes a one-line resume so the next station is not born amnesiac
 
 ```bash
-npm install 0xray@latest
+npm install 0xray
 ```
 
-xray is the pure v2 three-subsystem AI orchestration framework — **MCP-centric**, governed by Dynamo, and autonomous via thinDispatch. Consumer `npm install 0xray` auto-wires all four platform bridges and seven MCP servers with zero manual config.
+Consumer `npm install 0xray` auto-wires four platform bridges and seven MCP servers. Repertoire **0.2** is the factory memory organ (vendored): seed primitives, in-process routing, extra host MCP `repertoire`. Not an eighth `xray-*` server. Opt out with `"enabled": false, "provider": "repertoire"`.
 
 ## Quick Start
 
@@ -46,6 +48,7 @@ Docs: [guides/autonomy-command](docs-site/docs/guides/autonomy-command.md) · Sk
 
 | Version | Highlights |
 |---------|------------|
+| **4.0.0** | Exo + temperament. Constitution always on; ceremony by host. Repertoire enable-when-resolves + session resume. Survive the context window — still in development, not on npm yet. [vision](docs-site/docs/architecture/v4-vision.md) · [now](docs-site/docs/architecture/v4-now.md) |
 | **3.4.1** | Unified `install-bridges.cjs` on postinstall — OpenCode, Grok, Hermes, OpenClaw in one pass. All 7 MCPs servers via `npx -y 0xray mcp <cmd>` (no `dist/` paths). Canonical `release.mjs` pipeline. |
 | **3.3.1** | Orchestrator confidence gate wired into execution planning. |
 | **3.3.0** | Pluggable **Memory Routing** (`features.json` → `memory_routing`). Repertoire is the default provider in the framework repo. |
@@ -165,7 +168,7 @@ Pluggable `memory_routing` block in `features.json` (validated by `features.sche
 
 **External hosts** (Hermes/Grok): add `repertoire-mcp` to `.mcp.json` — see Repertoire docs.
 
-Consumers without Repertoire: `"memory_routing": { "enabled": false, "provider": "null" }`.
+Opt out: `"memory_routing": { "enabled": false, "provider": "repertoire" }`.
 
 Docs: [memory routing](docs-site/docs/guides/memory-routing.md) · [Repertoire](docs-site/docs/guides/repertoire.md) · [features.json](docs-site/docs/guides/features-json.md) · [all features since 3.1](docs-site/docs/guides/features-since-3.1.md)
 

package/dist/SKILLS.md

@@ -1,6 +1,6 @@
 # 0xray Skills
 
-**v3.5.3** — **45 skills** · invoked via `xray-skills` MCP · synced on postinstall
+**v4.0.0** — **45 skills** · invoked via `xray-skills` MCP · synced on postinstall
 
 xray ships 45 knowledge skills (`src/skills/<name>/SKILL.md`) across the full software lifecycle. Lead-dev operating model lives in **`orchestrator`** skill + `multi_agent_orchestration.lead_dev_mode` in features.json. Skills load on demand via the **xray-skills** MCP server (`npx -y 0xray mcp skills`), agent `@mentions`, or platform skill directories after postinstall.
 

package/dist/cli/commands/hermes-install.js

@@ -6,7 +6,8 @@ import { frameworkLogger } from '../../core/framework-logger.js';
 import { syncBuiltinSkills } from './skill-install.js';
 const __dirname = path.dirname(new URL(import.meta.url).pathname);
 const require = createRequire(import.meta.url);
-const wiring = require(path.join(__dirname, '..', '..', '..', 'scripts', 'node', 'bridge-mcp-wiring.cjs'));
+const packageRoot = path.join(__dirname, '..', '..', '..');
+const wiring = require(path.join(packageRoot, 'scripts', 'node', 'bridge-mcp-wiring.cjs'));
 export function registerHermesCommands(hermesCmd) {
     hermesCmd
         .command('install')
@@ -62,6 +63,10 @@ async function installForHermes(options = {}) {
             }
             frameworkLogger.log('hermes-integration', 'skills-synced', 'info', { count: skillsCopied });
         }
+        wiring.copyHermesFindProjectRootHelper(packageRoot, targetPluginDir);
+        if (wiring.copyHermesHookRuntimes(packageRoot)) {
+            frameworkLogger.log('hermes-integration', 'hooks-copied', 'info', {});
+        }
         const wired = wiring.wireHermesBridge(targetDir);
         console.log(`\x1b[32m✓ Wired Hermes mcp_servers (${wired.count} servers)\x1b[0m`);
         console.log(`\x1b[32m✓ Consumer root → ${targetDir}\x1b[0m`);

package/dist/cli/commands/openclaw-install.js

@@ -5,6 +5,7 @@ import { createRequire } from 'module';
 import { frameworkLogger } from '../../core/framework-logger.js';
 import { syncBuiltinSkills } from './skill-install.js';
 import { OpenClawConfigLoader } from '../../integrations/openclaw/config.js';
+import { writeSuitSessionBoot } from '../../nucleus/suit-temperament.js';
 const __dirname = path.dirname(new URL(import.meta.url).pathname);
 const require = createRequire(import.meta.url);
 const wiring = require(path.join(__dirname, '..', '..', '..', 'scripts', 'node', 'bridge-mcp-wiring.cjs'));
@@ -47,6 +48,19 @@ async function installForOpenClaw(options = {}) {
         const wired = wiring.wireOpenClawBridge(targetDir);
         console.log(`\x1b[32m✓ Wired OpenClaw MCP servers (${wired.count} via ${wired.method}) → ${wired.path}\x1b[0m`);
         console.log(`\x1b[32m✓ Consumer root → ${targetDir}\x1b[0m`);
+        try {
+            writeSuitSessionBoot(targetDir, 'openclaw', { source: '0xray/openclaw-install' });
+        }
+        catch {
+            /* session-boot is best-effort */
+        }
+        const packageRoot = path.resolve(__dirname, '..', '..', '..');
+        const hookInstalled = wiring.installOpenClawHostWear(packageRoot);
+        const cliBackend = wiring.maybeWriteOpenClawCliBackend();
+        frameworkLogger.log('openclaw-integration', 'cli-backend', 'info', { written: cliBackend });
+        if (hookInstalled) {
+            console.log(`\x1b[32m✓ OpenClaw PreToolUse hook → ${hookInstalled}\x1b[0m`);
+        }
         console.log('\n✅ 0xRay OpenClaw integration configured!');
         console.log('Run `openclaw mcp list` to verify MCP servers.');
         console.log('Edit `.xray/config/openclaw.json` for gateway URL, auth token, and device ID.');

package/dist/core/features-config.d.ts

@@ -7,6 +7,7 @@
  * @since 2026-01-25
  */
 import type { MemoryRoutingConfig } from "../memory-routing/types.js";
+import type { SuitHost, SuitProfile, SuitTemperamentConfig } from "../nucleus/suit-temperament.js";
 export type { MemoryRoutingConfig };
 export interface TokenOptimizationConfig {
     enabled: boolean;
@@ -63,16 +64,33 @@ export interface MultiAgentOrchestrationConfig {
     auto_consult_major_work?: boolean;
     /** Codex 69 — block new MCP/skill/handler files via PreToolUse (default true) */
     no_new_surface?: boolean;
-    /** 3.5.0 — PreToolUse gate after delegate-deferred (default true when lead_dev_mode) */
+    /** PreToolUse gate after delegate-deferred (default true when lead_dev_mode) */
     auto_chain_delegations?: boolean;
+    /** Hours before unstarted lead-dev plan is archived (default 8) */
+    plan_stale_hours?: number;
+    /** Hours spawn-plan-stale persists after archival (default 24) */
+    plan_archive_marker_hours?: number;
+    /** Sibling workspace roots for multi-repo session boot metadata */
+    sibling_repos?: Array<string | {
+        path: string;
+        label?: string;
+    }>;
+    /** Auto-run 3-agent confer at synthesis checkpoint */
+    confer_on_synthesis?: boolean;
+    confer?: {
+        enabled?: boolean;
+        on_synthesis?: boolean;
+    };
 }
+/** 0xRay 4.0 — nucleus SSOT. See docs-site/docs/guides/v3-temperament.md */
+export type { SuitHost, SuitProfile, SuitTemperamentConfig };
 /** Reflection triggers merged from legacy auto_reflection when synthesis is configured */
 export interface SynthesisReflectionConfig {
     mode?: AutoReflectionConfig['mode'];
     triggers?: AutoReflectionConfig['triggers'];
     thresholds?: AutoReflectionConfig['thresholds'];
 }
-/** 3.6.0 — periodic reflect & realign checkpoint (Synthesis PR1) */
+/** Periodic reflect & realign checkpoint (Synthesis) */
 export interface SynthesisConfig {
     enabled: boolean;
     every_n_gates: number;
@@ -384,10 +402,20 @@ export interface FeaturesConfig {
     auto_reflection?: AutoReflectionConfig;
     inference?: InferenceConfig;
     memory_routing?: MemoryRoutingConfig;
+    /** P2.2 — lightweight Grok write-tool postprocessor marker */
+    grok_postprocessor_light?: boolean;
+    /** P2 — git post-commit session inference capture */
+    inference_session_capture?: {
+        enabled?: boolean;
+        min_commits?: number;
+        lookback_commits?: number;
+    };
     synthesis: SynthesisConfig;
     kernel?: KernelConfig;
     processors?: ProcessorsConfig;
     enforcement?: EnforcementConfig;
+    /** v3 temperament — missing key means guided (existing consumers unchanged) */
+    suit_temperament?: SuitTemperamentConfig;
 }
 export type TaskType = "file_read" | "grep_search" | "simple_edit" | "bulk_refactor" | "architecture_review" | "security_audit" | "git_operations" | "documentation" | "code_generation" | "debugging" | "unknown";
 /**

package/dist/governance/governance-service.d.ts

@@ -29,6 +29,7 @@ export declare class GovernanceService {
      * Dynamo Solar SSOT is treated as a hard requirement by default.
      */
     govern(request: GovernanceRequest): Promise<GovernanceResponse>;
+    private resolveRequireExternalDynamo;
     private runGovernanceWithTimeout;
     private callSkillServer;
     private callExternalDynamo;