PabloTheThinker · PabloTheThinker · Aug 22, 2026 · Aug 3, 2026 · Aug 3, 2026 · Aug 3, 2026
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -0,0 +1,73 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ci-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  python:
+    name: Python ${{ matrix.python }}
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python: ["3.10", "3.12", "3.13"]
+    env:
+      HERMESPACE_NEURAL_VERBALIZE: "0"
+      HERMESPACE_SKIP_NEURAL: "1"
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python }}
+          cache: pip
+      - name: Install clean package
+        run: python -m pip install .
+      - name: Unit tests
+        run: python -m unittest discover -s tests -v
+      - name: Hermes v0.20 host contract
+        run: python scripts/verify_hermes_integration.py
+
+  integration:
+    name: Smoke and operational E2E
+    runs-on: ubuntu-latest
+    env:
+      HERMESPACE_NEURAL_VERBALIZE: "0"
+      HERMESPACE_SKIP_NEURAL: "1"
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: pip
+      - run: python -m pip install .
+      - name: Security audit
+        run: ./scripts/security_audit.sh
+      - name: Lint production additions
+        run: |
+          python -m pip install ruff
+          python -m ruff check --isolated \
+            src/hermespace/atomic.py \
+            src/hermespace/hermes_runtime.py \
+            src/hermespace/plugin.py \
+            scripts/verify_hermes_integration.py \
+            tests/test_plugin_contract.py \
+            tests/test_store.py
+      - name: Smoke
+        run: ./scripts/smoke_test.sh
+      - name: Ops end-to-end
+        run: ./scripts/e2e_ops.sh
+      - name: Week-one bench
+        run: python scripts/bench_week1.py
+      - name: Build wheel and sdist
+        run: |
+          python -m pip install build
+          python -m build
diff --git a/ABOUT.md b/ABOUT.md
@@ -1,4 +1,4 @@
-**Hermespace is a functional J-Space workbench for Hermes agents** — FOA desk, dual decode, and an append-only world projection. Not [J-Space](https://github.com/anomalyco/j-space) neural weights. Not a second runtime. A harness-level global workspace inside Hermes that runs **standalone** and is **powered by [HermesCube](https://github.com/PabloTheThinker/hermescube)** when present.
+**Hermespace is a functional Access Workspace workbench for Hermes agents** — FOA desk, dual decode, and an append-only world projection. Not [Access Workspace](https://github.com/anomalyco/j-space) neural weights. Not a second runtime. A harness-level global workspace inside Hermes that runs **standalone** and is **powered by [HermesCube](https://github.com/PabloTheThinker/hermescube)** when present.
 
 Alongside the desk: skills+MEMORY fabric, neural FOA, autonomy grid, Cube heart/center cable (soft-fail).
 
@@ -38,7 +38,7 @@ Hermespace solves this with two systems:
 | Project | Relation |
 |---------|----------|
 | [ActiveGraph](https://github.com/anomalyco/ActiveGraph) | Event log is the agent, graph is the world — influenced the archive-first design |
-| [J-Space](https://github.com/anomalyco/j-space) | Verbal workspace, ~25 concepts, broadcasting hub — inspired the `concepts` system and `_refresh_concepts()` |
+| [Access Workspace](https://github.com/anomalyco/j-space) | Verbal workspace, ~25 concepts, broadcasting hub — inspired the `concepts` system and `_refresh_concepts()` |
 | [Missing Knowledge Layer](https://arxiv.org/abs/2405.10697) | Knowledge = supersession, Memory = decay, Wisdom = evidence-gated — inspired the epoch progression from unfiltered archive → evidence-gated wisdom |
 | [Vygotsky](https://en.wikipedia.org/wiki/Inner_speech) | Inner speech, signs as instruments of thought — influenced the dual decode architecture (report to human vs context to model) |
 | [Hermes Agent](https://github.com/NousResearch/hermes-agent) | The agent that grows with you, by Nous Research |

diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
@@ -33,7 +33,8 @@ PYTHONPATH=src python3 -m unittest discover -s tests -v
 | Agent skill | `skills/hermespace/` |
 | CLI | `src/hermespace/cli.py`, `scripts/hs` |
 | Tests | `tests/` |
-| North star | `PURPOSE.md`, `docs/CODEMAP.md` |
+| North star | `PURPOSE.md`, `LAYOUT.md`, `docs/architecture/CODEMAP.md` |
+| Assessment | `docs/assessment/28-hermes-agent-jspace-assessment.md` |
 | Maintainer note | `FOR_HERMES.md` |
 | Integration doors | `INTEGRATION.md`, `WORKFLOW.md` |
 

diff --git a/FOR_HERMES.md b/FOR_HERMES.md
@@ -1,89 +1,3 @@
-# For Hermes Agent maintainers & dogfooders
+# FOR_HERMES
 
-**Repo:** https://github.com/PabloTheThinker/hermespace  
-**Companion to:** [NousResearch/hermes-agent](https://github.com/NousResearch/hermes-agent)  
-**Not** an official Nous product — independent open companion workspace.
-
-## What this is (30 seconds)
-
-Hermespace adds a **limited working-memory desk** beside Hermes:
-
-1. **Workbench** — FOA, park stack, load-aware `receive_order`
-2. **Dual decode** — short human `report` vs dense model `context` (never dump inject to chat)
-3. **Fabric** — rank *this* `$HERMES_HOME/skills` + inject MEMORY/USER excerpts
-4. **Plugin** — `on_session_start` / `pre_llm_call` / `on_session_end` broadcast when desk ready
-5. **Study DB** — turns under `~/.hermespace` (local; not Hermes session DB)
-
-It does **not** replace Hermes tools, skills, memory, gateway, or the agent loop.
-
-## Fast path
-
-```bash
-git clone https://github.com/PabloTheThinker/hermespace.git
-cd hermespace
-./scripts/install_hermes.sh
-./scripts/smoke_test.sh    # expect 9/9
-```
-
-Then in any agent session:
-
-```python
-from hermespace import Workbench
-wb = Workbench(agent_id="hermes", session_id="main")
-wb.enter()
-r = wb.receive_order("…", goal="…", say="…", force=True)
-# r["user_reply"] → user   r["model_context"] → model
-```
-
-Or CLI: `./scripts/hs turn -m "…" --goal "…" --say "…" --force`
-
-## Design honesty
-
-| Claim | Reality |
-|-------|---------|
-| “Like Claude J-space” | **Role** only — limited workspace *outside* weights. No activation access. |
-| Consciousness / brain scan | **No.** Harness cognition + optional local embeddings. |
-| Replaces Hermes skills/memory | **No.** Ranks and injects them. |
-| Official Nous module | **No.** Community companion; feedback welcome. |
-
-Inspired by working-memory limits (Baddeley-style capacity, load, GWT broadcast as *metaphors* for agent UX) and Anthropic’s public J-space *research framing* — implemented as open Hermes integration, not a closed-model probe.
-
-## Integration surface
-
-| Door | Entry |
-|------|--------|
-| Plugin | `hermes_plugin/` → `$HERMES_HOME/plugins/hermespace` |
-| Skill | `skills/hermespace/` → `$HERMES_HOME/skills/hermespace` |
-| Python | `hermespace.agent_api`, `hermespace.Workbench` |
-| CLI | `scripts/hs` |
-| Smoke | `scripts/smoke_test.sh` |
-
-Hook implementation lives in `src/hermespace/hermes_bridge.py` (plugin is thin `register`).
-
-When the plugin directory is a **symlink into this checkout**, import auto-resolves `../src` so dogfooders often need no `HERMESPACE_ROOT`.
-
-## Autonomy grid (v0.14)
-
-Missions, lenses, dream, self-talk, skillbench (hot-swap / merge / mutate), title tree.  
-Ground-up for Hermespace — not a port of AgentDrive or Conductor.
-
-```bash
-hs grid status
-hs grid dream --force
-```
-
-See docs/18-autonomy-grid.md. Autonomy self-order stays **off** unless `HERMESPACE_AUTONOMY=1`.
-
-## What we want from Hermes dogfood
-
-- Does plugin register cleanly on current Hermes?
-- Is dual-decode useful on Telegram/gateway, or noisy?
-- Fabric skill ranking quality on real skill trees?
-- Gaps vs native Hermes memory / Honcho / skills loop?
-- Would you want any of this upstream-shaped (skill only, plugin only, or neither)?
-
-Open issues/PRs on the repo. Thanks for looking.
-
-## License
-
-MIT — see `LICENSE`. Hermes Agent remains Nous Research’s project under its own license.
+Moved to **[docs/integration/FOR_HERMES.md](docs/integration/FOR_HERMES.md)**.
diff --git a/INTEGRATION.md b/INTEGRATION.md
@@ -22,7 +22,7 @@ hermes plugins enable hermespace
 ```
 
 Agent skill SoT: [`skills/hermespace/SKILL.md`](skills/hermespace/SKILL.md).  
-Recommended env: see `docs/RECOMMENDED.md`.
+Recommended env: see `docs/ops/RECOMMENDED.md`. · Layout: `LAYOUT.md`.
 
 ## Door A — Python (best for agents)
 
@@ -122,12 +122,15 @@ send_user(result["user_reply"])
 
 See docs/14-workbench-pocket-dimension.md.
 
-## Hermes framework hooks (v0.12)
+## Hermes framework hooks (v0.20+)
 
 Plugin registers:
-- `on_session_start` — workbench enter + env kit context
-- `pre_llm_call` — desk + neural broadcast + workbench footer
-- `on_session_end` — idle_tick
+- `on_session_start` — initialize the session scope
+- `pre_llm_call` — bounded ephemeral Access Workspace context
+- `post_llm_call` / `post_tool_call` — observe completed work
+- `on_session_end` — lightweight turn boundary
+- `on_session_finalize` — final harvest + idle maintenance
+- `on_session_reset` / `subagent_start` / `subagent_stop` — gateway/A2A lifecycle
 
 Why: docs/16-why-hermes-framework.md
 

diff --git a/LAYOUT.md b/LAYOUT.md
@@ -0,0 +1,60 @@
+# Hermespace codespace layout
+
+North star: [PURPOSE.md](PURPOSE.md) · Assessments:
+[28-hermes-agent-jspace](docs/assessment/28-hermes-agent-jspace-assessment.md) ·
+[33-living-memories-cube-world](docs/assessment/33-living-memories-cube-world.md)
+
+## Repository map
+
+```
+hermespace/
+├── PURPOSE.md / ABOUT.md / README.md / LAYOUT.md
+├── src/hermespace/          # Python package
+│   ├── access/              # ★ Access Engine (hub · env · protocol · engine)
+│   ├── turn/                # planned — workflow/desk/gate (README only)
+│   ├── memory/              # planned — world/episodic/semantic (README only)
+│   ├── warehouse/           # planned — cube cable (README only)
+│   ├── grid/                # autonomy grid (missions · dream · skillbench)
+│   ├── hermes_bridge.py     # Hermes plugin hooks
+│   └── …                    # turn spine still at package root (stable imports)
+├── hermes_plugin/           # thin Hermes register()
+├── desktop_plugin/          # Hermes Desktop page/pane
+├── skills/hermespace/       # agent skill
+├── docs/
+│   ├── access/              # Access Workspace map · thesis · environment
+│   ├── assessment/          # deep assessments
+│   ├── architecture/        # CODEMAP · Cube contract
+│   ├── integration/         # Hermes fit · FOR_HERMES
+│   ├── ops/                 # pulse · everyday · recommended
+│   ├── research/            # numbered research archive
+│   └── roadmap/             # open backlog · OEW phases
+├── tests/ · scripts/ · experiments/ · benchmarks/
+├── runtime/ · spec/         # templates / protocol specs
+└── assets/
+```
+
+## Layer authority
+
+| Layer | Path | Authority |
+|-------|------|-----------|
+| OEW / Access Workspace | `src/hermespace/access/` | Turn FOA + audit SoT |
+| Turn spine | package root (`workflow`, `desk`, …) | Live turn |
+| Autonomy | `grid/` | Missions / dream / skillbench |
+| Identity projection | `world.py` … | Recharged from Cube |
+| Warehouse cable | `cube_module.py` | Soft-fail Cube heart |
+| Hermes surface | `hermes_plugin/` + bridge | Hooks only |
+
+## Import rules
+
+Prefer:
+
+```python
+from hermespace.jspace import JSpace, JSpaceEnv, evaluate_material_turn
+```
+
+Compat: `hermespace.jspace_env` still re-exports `JSpaceEnv`.
+
+## Moving code later
+
+Planned packages under `turn/`, `memory/`, `warehouse/` hold READMEs only until
+OEW Phase B is green — avoid churning imports while the protocol lands.