Obsidian-Owl · MacAttak · Feb 15, 2026 · Feb 15, 2026
diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -67,7 +67,7 @@ jobs:
 
       - name: Update README.md
         run: |
-          sed -i "s/^\*\*Version\*\*: .*/**Version**: ${{ steps.next.outputs.version }}/" README.md
+          sed -i "s/MIT License · v[0-9]*\.[0-9]*\.[0-9]*/MIT License · v${{ steps.next.outputs.version }}/" README.md
 
       - name: Update banner version
         run: |
@@ -79,20 +79,26 @@ jobs:
           NEW_VERSION="${{ steps.next.outputs.version }}"
           BODY=$(cat /tmp/changelog_body.md)
 
-          # Create new section
-          {
-            # Copy header (lines before first ## [)
-            sed -n '1,/^## \[/{ /^## \[/!p }' CHANGELOG.md
-            echo ""
-            echo "## [${NEW_VERSION}] - ${DATE}"
-            echo ""
-            echo "### Changed"
-            echo ""
-            echo "$BODY"
-            echo ""
-            # Copy rest of file starting from first ## [
-            sed -n '/^## \[/,$p' CHANGELOG.md
-          } > CHANGELOG.tmp && mv CHANGELOG.tmp CHANGELOG.md
+          # Skip if a hand-written entry already exists for this version
+          if grep -q "^## \[${NEW_VERSION}\]" CHANGELOG.md; then
+            echo "CHANGELOG already contains entry for ${NEW_VERSION}, updating date only"
+            sed -i "s/^## \[${NEW_VERSION}\].*/## [${NEW_VERSION}] - ${DATE}/" CHANGELOG.md
+          else
+            # Create new section
+            {
+              # Copy header (lines before first ## [)
+              sed -n '1,/^## \[/{ /^## \[/!p }' CHANGELOG.md
+              echo ""
+              echo "## [${NEW_VERSION}] - ${DATE}"
+              echo ""
+              echo "### Changed"
+              echo ""
+              echo "$BODY"
+              echo ""
+              # Copy rest of file starting from first ## [
+              sed -n '/^## \[/,$p' CHANGELOG.md
+            } > CHANGELOG.tmp && mv CHANGELOG.tmp CHANGELOG.md
+          fi
 
       - name: Create release branch and PR
         env:

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -8,14 +8,31 @@ Versions follow [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 
 
+## [0.7.0] - 2026-02-16
 
+### Added
+
+- **Codebase Landscape** (`LANDSCAPE.md`) — Persistent codebase knowledge document covering architecture, modules, conventions, and gotchas. Created during `/sw-init` (optional survey phase), loaded by `/sw-design` with automatic staleness refresh, incrementally updated by `/sw-learn` after shipping. New `protocols/landscape.md` defines format, size targets (500-3000 words), and freshness rules (7-day default, configurable). (#26)
+- **Codebase Audit** (`/sw-audit`) — Periodic health check skill that finds systemic issues per-change gates miss. Four dimensions: architecture, complexity, consistency, debt. Delegates to architect and reviewer agents, synthesizes findings into persistent `AUDIT.md` with stable finding IDs across re-runs. Adaptive intensity: focused (path), standard (small projects), full (large projects, parallel agents). New `protocols/audit.md` defines finding format, ID matching, lifecycle, and freshness. (#27)
+- **Audit integration with design** — `/sw-design` now loads `AUDIT.md` during research and surfaces relevant findings for the area being designed
+- **Audit integration with learn** — `/sw-learn` checks if shipped work addresses open audit findings and marks them resolved
 
+### Changed
+
+- `protocols/context.md` — Reference Documents section now includes both `LANDSCAPE.md` and `AUDIT.md` (optional, load on demand)
+- Skill count: 14 → 15 (10 user-facing + 5 gates)
+- Protocol count: 11 → 12
+- README updated with Codebase Knowledge and Health section, version footer corrected
+- DESIGN.md Reference Documents section documents both LANDSCAPE.md and AUDIT.md lifecycles
 
 ## [0.6.0] - 2026-02-15
 
-### Changed
+### Added
 
-- feat(skills): adaptive intensity, session hooks, stop guard, compaction recovery (#23)
+- **Adaptive intensity** — `/sw-design` triages requests as Full, Lite, or Quick based on complexity. Lite skips design.md, Quick skips straight to build
+- **Session hooks** — `hooks/session-start.md` loads recovery protocol on compaction detection
+- **Stop guard** — `/sw-guard` can now configure stop-on-error hooks for CI/CD layers
+- **Compaction recovery** — All stateful skills check for compaction and resume from last completed step
 
 ## [0.5.2] - 2026-02-13
 
@@ -35,25 +52,6 @@ Versions follow [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [0.5.0] - 2026-02-12
 
-### Changed
-
-- fix(docs): address stale wording from wiring gate findings
-- docs: update README and CHANGELOG for design/plan split
-- docs: update DESIGN.md and CLAUDE.md for design/plan split
-- feat(skills): update sw-verify and sw-status for design/plan split
-- feat(skills): add design.md to sw-build inputs and context envelope
-- feat(skills): rewrite sw-plan for decomposition and specs only
-- fix(protocols): add null-to-designing transition in state table
-- feat(skills): create sw-design skill for interactive solution architecture
-- feat(protocols): add designing status and sw-design handoff
-- fix(agents): correct executor description, add tester negative constraints
-- docs(design): add behavioral discipline as principle #7
-- feat(sw-build): scope REFACTOR step and add behavioral envelope
-- feat(sw-plan): add simplicity bias to Design phase
-- feat(agents): add Karpathy-aligned behavioral discipline sections
-
-## [0.5.0] - 2026-02-13
-
 ### Added
 
 - **sw-design skill** — Interactive solution architecture with research, adversarial critic, and user approval. Produces `design.md`, `context.md`, and conditional artifacts (`decisions.md`, `data-model.md`, `contracts.md`, `testing-strategy.md`, `infra.md`, `migrations.md`)

diff --git a/README.md b/README.md
@@ -34,6 +34,8 @@ Specwright closes the **entire loop** — design, plan, build, verify, ship, lea
 - Evidence-based PRs with gate proof for every acceptance criterion
 - Compaction recovery reloads full context automatically
 - Learning system captures failures and promotes patterns across sessions
+- Codebase knowledge persists across sessions — no re-discovering the same architecture
+- Periodic health checks find systemic debt that per-change gates miss
 - One install, configure once, works with any language or framework
 
 ## What Makes This Different
@@ -48,6 +50,10 @@ Specwright focuses on the **verification and evidence** side — the part where
 
 **Learning System** — Failures are captured, patterns are promoted, and learnings compact into tiered memory (index, themes, raw data). The system gets smarter with every session. Knowledge survives context windows.
 
+**Codebase Knowledge** — During init, Specwright surveys your codebase and builds a persistent knowledge document (`LANDSCAPE.md`) covering architecture, modules, conventions, and gotchas. Design phases load this instantly instead of re-scanning. It stays current — refreshed when stale, incrementally updated after every shipped work unit.
+
+**Codebase Health Checks** — Run `/sw-audit` periodically to find systemic issues that per-change gates miss: architecture debt, complexity growth, convention drift, accumulated workarounds. Findings persist in `AUDIT.md` with stable IDs across re-runs. Design phases surface relevant findings. The learn phase resolves them when addressed.
+
 **Compaction Recovery** — All stateful skills support resume-from-crash. When Claude's context window compacts, Specwright reloads full state from disk. The only plugin in this space that handles context loss gracefully.
 
 ## How It Works
@@ -62,6 +68,7 @@ graph LR
     F -.->|next work unit| D
     F --> G["/sw-learn"]
     G -.->|patterns feed back| B
+    H["/sw-audit"] -.->|findings feed into| B
 
     style A fill:#1e293b,stroke:#f59e0b,color:#f8fafc
     style B fill:#1e293b,stroke:#f59e0b,color:#f8fafc
@@ -70,6 +77,7 @@ graph LR
     style E fill:#1e293b,stroke:#f59e0b,color:#f8fafc
     style F fill:#1e293b,stroke:#f59e0b,color:#f8fafc
     style G fill:#1e293b,stroke:#f59e0b,color:#f8fafc
+    style H fill:#1e293b,stroke:#f59e0b,color:#f8fafc
 ```
 
 | Phase | What Happens | Key Innovation |
@@ -81,6 +89,7 @@ graph LR
 | **Verify** | 5 quality gates with evidence capture | Findings shown inline, not just pass/fail badges |
 | **Ship** | PR with acceptance criteria mapped to evidence | Every requirement traceable to code + test |
 | **Learn** | Capture patterns, promote to constitution | Knowledge compounds across sessions |
+| **Audit** | Periodic health check — architecture, complexity, consistency, debt | Finds systemic issues gates miss. Run anytime. |
 
 ## Quick Start
 
@@ -111,6 +120,30 @@ Then design, plan, and iterate per work unit:
 /sw-ship
 ```
 
+## Codebase Knowledge and Health
+
+Two optional features keep Specwright informed about your codebase across sessions:
+
+**Landscape** (`LANDSCAPE.md`) — A persistent map of your codebase's architecture, modules, conventions, and integration points. Created automatically during `/sw-init` if you opt in. The design phase loads it for instant context instead of re-scanning every time. Updated incrementally after each shipped work unit.
+
+- Created by: `/sw-init` (survey phase, optional)
+- Consumed by: `/sw-design` (auto-refreshed when stale)
+- Updated by: `/sw-learn` (after shipping)
+
+**Audit** (`AUDIT.md`) — A persistent record of systemic codebase health issues. Run `/sw-audit` when you want a health check — it's not part of the regular workflow, so use it whenever it makes sense: before starting a large feature, after a refactoring sprint, or on a regular cadence.
+
+```
+/sw-audit              # auto-triage: standard or full based on codebase size
+/sw-audit src/api/     # focused: analyze only the specified path
+/sw-audit --full       # full: parallel analysis across all dimensions
+```
+
+Findings persist across runs with stable IDs. When you design new work, relevant findings are surfaced automatically. When you ship work that addresses a finding, the learn phase marks it resolved.
+
+- Created by: `/sw-audit` (run anytime)
+- Consumed by: `/sw-design` (surfaces relevant findings during research)
+- Resolved by: `/sw-learn` (marks addressed findings as resolved)
+
 ## Six Specialized Agents
 
 Specwright delegates to purpose-built agents — each with a distinct role, model, and adversarial stance:
@@ -136,14 +169,20 @@ Every work unit passes through configurable gates before shipping. **Default sta
 | **Wiring** | Orphaned files, unused exports, layer violations, circular deps | WARN |
 | **Spec** | Every acceptance criterion mapped to code + test evidence | BLOCK |
 
-## Anchor Documents
+## Persistent Documents
 
-Two persistent documents drive all decisions and survive context compaction:
+Two **anchor documents** drive all decisions and survive context compaction:
 
 **`CONSTITUTION.md`** — Development practices the AI must follow. Testing standards, coding conventions, security requirements. Not suggestions — rules.
 
 **`CHARTER.md`** — Technology vision and architectural invariants. What this project is, who consumes it, what doesn't change.
 
+Two optional **reference documents** accelerate research and track health:
+
+**`LANDSCAPE.md`** — Codebase knowledge: architecture, modules, conventions, gotchas. Loaded on demand, never blocks workflow.
+
+**`AUDIT.md`** — Codebase health findings: systemic debt, complexity growth, convention drift. Loaded on demand, findings have stable IDs.
+
 ## Skills
 
 <table>
@@ -218,5 +257,5 @@ Specwright is open source under the MIT license.
 ---
 
 <p align="center">
-  <sub>Built by <a href="https://github.com/Obsidian-Owl">ObsidianOwl</a> · MIT License · v0.5.0</sub>
+  <sub>Built by <a href="https://github.com/Obsidian-Owl">ObsidianOwl</a> · MIT License · v0.6.0</sub>
 </p>