diff --git a/.cursorrules b/.cursorrules
new file mode 100644
index 0000000..e0c37ff
--- /dev/null
+++ b/.cursorrules
@@ -0,0 +1,45 @@
+# KyaniteLabs — AI Agent Instructions
+
+## Organization
+- Org: KyaniteLabs, owner: Pastorsimon1798
+- All repos use issue-driven development — contributions enter as GitHub issues, not direct PRs
+- Pipeline runs every 30 min: triage → fix → review → merge
+- All main/master branches are protected (no direct push, CI required)
+
+## Code Quality
+- Formatter: ruff (Python), prettier (JS/TS)
+- Linter: ruff (Python), eslint (JS/TS)
+- Config: pyproject.toml (Python), pnpm (Node)
+- Pin exact dependency versions
+- No orphaned code — every function called, every file imported
+- No commented-out code blocks
+
+## Security
+- No `shell=True` with user input
+- No hardcoded secrets/tokens/API keys
+- Validate file paths (no path traversal)
+- Use HTTPS for external calls
+- Parameterized queries for databases
+
+## Testing
+- Test behavior, not implementation
+- Test the unhappy path (API down, file missing, bad input)
+- Integration tests over mocks for external systems
+- One assertion per concept
+
+## Git
+- Commits tell why, not what
+- PRs under 400 lines
+- Rebase, don't merge on feature branches
+- Never skip CI
+
+## Local LLM
+- Use local inference at 100.66.225.85:1234 before cloud APIs
+- Check loaded models first, don't touch models you didn't load
+- Unload when done
+- CPU thread pool: 10, flash attention: on, KV cache: Q8
+
+## Agent Coordination
+- Check open PRs before editing files
+- Leave context in issues, not in code
+- Run CI checks locally before pushing
diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md
new file mode 100644
index 0000000..e0c37ff
--- /dev/null
+++ b/.github/copilot-instructions.md
@@ -0,0 +1,45 @@
+# KyaniteLabs — AI Agent Instructions
+
+## Organization
+- Org: KyaniteLabs, owner: Pastorsimon1798
+- All repos use issue-driven development — contributions enter as GitHub issues, not direct PRs
+- Pipeline runs every 30 min: triage → fix → review → merge
+- All main/master branches are protected (no direct push, CI required)
+
+## Code Quality
+- Formatter: ruff (Python), prettier (JS/TS)
+- Linter: ruff (Python), eslint (JS/TS)
+- Config: pyproject.toml (Python), pnpm (Node)
+- Pin exact dependency versions
+- No orphaned code — every function called, every file imported
+- No commented-out code blocks
+
+## Security
+- No `shell=True` with user input
+- No hardcoded secrets/tokens/API keys
+- Validate file paths (no path traversal)
+- Use HTTPS for external calls
+- Parameterized queries for databases
+
+## Testing
+- Test behavior, not implementation
+- Test the unhappy path (API down, file missing, bad input)
+- Integration tests over mocks for external systems
+- One assertion per concept
+
+## Git
+- Commits tell why, not what
+- PRs under 400 lines
+- Rebase, don't merge on feature branches
+- Never skip CI
+
+## Local LLM
+- Use local inference at 100.66.225.85:1234 before cloud APIs
+- Check loaded models first, don't touch models you didn't load
+- Unload when done
+- CPU thread pool: 10, flash attention: on, KV cache: Q8
+
+## Agent Coordination
+- Check open PRs before editing files
+- Leave context in issues, not in code
+- Run CI checks locally before pushing
diff --git a/.windsurfrules b/.windsurfrules
new file mode 100644
index 0000000..e0c37ff
--- /dev/null
+++ b/.windsurfrules
@@ -0,0 +1,45 @@
+# KyaniteLabs — AI Agent Instructions
+
+## Organization
+- Org: KyaniteLabs, owner: Pastorsimon1798
+- All repos use issue-driven development — contributions enter as GitHub issues, not direct PRs
+- Pipeline runs every 30 min: triage → fix → review → merge
+- All main/master branches are protected (no direct push, CI required)
+
+## Code Quality
+- Formatter: ruff (Python), prettier (JS/TS)
+- Linter: ruff (Python), eslint (JS/TS)
+- Config: pyproject.toml (Python), pnpm (Node)
+- Pin exact dependency versions
+- No orphaned code — every function called, every file imported
+- No commented-out code blocks
+
+## Security
+- No `shell=True` with user input
+- No hardcoded secrets/tokens/API keys
+- Validate file paths (no path traversal)
+- Use HTTPS for external calls
+- Parameterized queries for databases
+
+## Testing
+- Test behavior, not implementation
+- Test the unhappy path (API down, file missing, bad input)
+- Integration tests over mocks for external systems
+- One assertion per concept
+
+## Git
+- Commits tell why, not what
+- PRs under 400 lines
+- Rebase, don't merge on feature branches
+- Never skip CI
+
+## Local LLM
+- Use local inference at 100.66.225.85:1234 before cloud APIs
+- Check loaded models first, don't touch models you didn't load
+- Unload when done
+- CPU thread pool: 10, flash attention: on, KV cache: Q8
+
+## Agent Coordination
+- Check open PRs before editing files
+- Leave context in issues, not in code
+- Run CI checks locally before pushing
diff --git a/AGENTS.md b/AGENTS.md
new file mode 100644
index 0000000..812ba9b
--- /dev/null
+++ b/AGENTS.md
@@ -0,0 +1,302 @@
+# KyaniteLabs Engineering Rules
+
+**Repository**: devarch-framework
+**Description**: Forensic repository archaeology framework for git history analysis, development signals, and engineering narrative reports.
+**Tech Stack**: Python
+
+This file provides operating instructions for all AI coding agents working in KyaniteLabs repositories. It is the AGENTS.md mirror of CLAUDE.md — both files contain identical rules.
+
+See KyaniteLabs/.github for org-wide rules. Project-specific rules below:
+
+## Project-Specific Context
+
+This is the **devarch-framework** repository. It is Forensic repository archaeology framework for git history analysis, development signals, and engineering narrative reports..
+
+# KyaniteLabs Engineering Rules
+
+This file provides operating instructions for all AI coding agents working in KyaniteLabs repositories. It is the AGENTS.md mirror of CLAUDE.md — both files contain identical rules.
+
+## Organization Principles
+
+- **Local-first when it matters** — keep data, creative work, and knowledge close to the people who own it
+- **Useful over flashy** — solve real workflow problems
+- **Human-in-the-loop by design** — AI accelerates judgment, it doesn't erase it
+- **Open where it helps** — public projects are documented for discovery and reuse
+- **Craft matters** — README quality, tests, and maintenance are part of the product
+
+## Issue-Driven Development
+
+All contributions to KyaniteLabs repos flow through GitHub issues.
+
+### Workflow
+
+```
+External contributor opens issue
+  → Template applies: bug/enhancement + needs-triage label
+  → Maintainer reviews, adds: approved label
+  → Auto-triage promotes to: agent-ready label
+  → Pipeline picks up → creates fix PR
+
+Owner/member (Pastorsimon1798) opens issue
+  → Auto-triage detects member author
+  → Promotes directly to agent-ready (skips triage)
+  → Pipeline picks up → creates fix PR
+```
+
+### Label System
+
+| Label | Meaning | Who sets it |
+|-------|---------|------------|
+| `needs-triage` | New issue, awaiting review | Issue template (auto) |
+| `approved` | Maintainer approved for work | Maintainer (manual) |
+| `agent-ready` | Ready for pipeline to pick up | Auto-triage (auto) |
+| `bug` | Bug report | Issue template (auto) |
+| `enhancement` | Feature request | Issue template (auto) |
+| `repo-pipeline` | Created by pipeline scanner | Pipeline (auto) |
+| `ci-failure` | CI broke on main branch | Pipeline (auto) |
+| `severity:low/medium/high/critical` | Impact level | Pipeline (auto) |
+
+### Rules for Agents Creating Issues
+
+- Never create issues without the `repo-pipeline` label
+- Always include reproduction steps or evidence
+- Set severity based on: does it break main? does it affect users? is it cosmetic?
+- Link to the relevant CI run, commit, or file
+
+## Pipeline Awareness
+
+The GITHUB_pipeline runs a triage cycle every 30 minutes. When working in KyaniteLabs repos, be aware:
+
+1. **Your PRs will be monitored** by `pr-maintainer.py` — it fixes CI failures, rebases, and addresses review comments
+2. **Do not delete `fix/issue-*` branches** — the pipeline manages these
+3. **Pipeline-created issues** have the `repo-pipeline` label — treat them as real issues
+4. **Circuit breaker** — if the pipeline enters RED state, all automated work stops until manually reset
+5. **Issue priority** — pipeline sorts by severity (critical > high > medium > low), then by age
+
+## CI Standards
+
+All KyaniteLabs repos must have these CI checks (via GitHub Actions on Blacksmith runners):
+
+### Required Checks (must pass before merge)
+
+| Check | Tool | Config |
+|-------|------|--------|
+| Lint | ruff | `ruff check . && ruff format --check .` |
+| Test | pytest | `pytest --tb=short -q` |
+| Build | pip install | `pip install -e .` (Python) or equivalent |
+
+### Recommended Checks (add when applicable)
+
+| Check | Tool | When |
+|-------|------|------|
+| Security scan | bandit + pip-audit | Python repos |
+| Type check | mypy | repos with type annotations |
+| Docker build | docker build | repos with Dockerfile |
+| Package surface | custom script | published packages |
+| Compatibility matrix | multi-Python/Node | public packages |
+
+### CI Rules
+
+- All CI runs on `blacksmith-2vcpu-ubuntu-2404` runners
+- CI must pass on all supported Python/Node versions
+- Never skip CI (`--no-verify`) — if it fails, fix the root cause
+- CI failures on main get auto-detected by the pipeline
+
+## Branch Protection
+
+All `main`/`master` branches in KyaniteLabs repos are protected:
+
+- No direct pushes — everything through PRs
+- Required status checks must pass
+- Branches must be up-to-date before merge
+- Force pushes blocked
+- Linear history required (squash or rebase merge)
+
+## Repository Standards
+
+### Python Repos
+- **Formatter**: ruff (not black)
+- **Linter**: ruff
+- **Config**: `pyproject.toml` (not setup.py/setup.cfg)
+- **Python**: minimum 3.11, test on 3.11 and latest
+- **Dependencies**: pin exact versions
+
+### Node/TypeScript Repos
+- **Package manager**: pnpm (not npm/yarn)
+- **Formatter**: prettier
+- **Linter**: eslint
+
+### All Repos
+- **README.md**: Must explain what it is, how to run it, how to test it
+- **CHANGELOG.md**: Updated with every release
+- **LICENSE**: MIT (org default)
+- **CONTRIBUTING.md**: Points to org contributing guide
+
+## Agent Coordination
+
+When multiple agents work on the same repo:
+
+1. **Claim your scope** — note in the issue which files/areas you're working on
+2. **Don't edit files another agent is actively modifying** — check open PRs first
+3. **Rebase frequently** — pull from main before pushing to avoid conflicts
+4. **Test before pushing** — run the same CI checks locally first
+5. **Communicate via issues** — leave comments on issues for context, not in code
+
+## Code Quality Rules
+
+### Dependency Discipline
+- Never add a dependency for something 20 lines of stdlib can do
+- Never add a dependency without checking it's actively maintained (commits in last 6 months)
+- Pin exact versions in requirements, never ranges
+- When you add one, write a one-line comment explaining why
+
+### Error Handling
+- Validate at system edges only (user input, API responses, file reads)
+- Trust internal code — no try/except around your own functions unless the operation is genuinely fallible (network, disk)
+- Never catch `Exception` broadly — catch the specific failure mode
+- Errors must be actionable: tell the human *what* broke and *what to do*
+
+### No Orphaned Code
+- Every function must be called by something (tests count)
+- Every file must be imported by something
+- No dead config, no unused imports, no commented-out code blocks
+- If you write it, wire it up or delete it
+
+### Security Non-Negotiables
+- Never use `shell=True` with user input
+- Never hardcode secrets, tokens, or API keys
+- Always validate file paths (no path traversal)
+- Always use parameterized queries for database interaction
+- Always use HTTPS for external calls
+- Sanitize any string that becomes part of a command, HTML, or SQL
+
+### Testing
+- Test behavior, not implementation
+- One assertion per concept
+- Test the unhappy path — what happens when the API is down, the file is missing, the input is garbage
+- Integration tests over unit tests for anything touching external systems
+- No mocking what you don't own unless the real thing is destructive or slow
+
+### Documentation
+- README must answer: what is this, how do I run it, how do I test it
+- Public functions get a docstring only if the name doesn't explain itself
+- No docstrings on private functions — rename them instead
+- CHANGELOG entries for every user-visible change
+
+### Git and PR Hygiene
+- Commits tell *why*, not *what* (the diff shows what)
+- PRs under 400 lines of meaningful change
+- Every PR describes how it was verified
+- No merge commits on feature branches — rebase
+
+### Performance
+- No N+1 queries
+- No loading full datasets into memory — paginate or stream
+- No regex in hot loops
+- If you add a cache, also add a cache invalidation strategy
+
+### Configuration
+- Magic numbers become named constants
+- Tunable behavior becomes config/env vars, not code changes
+- Default config works for development without modification
+
+## KyaniteLabs Tech Stack Standard
+
+Every project in KyaniteLabs follows these stack decisions. No exceptions without explicit approval.
+
+### TypeScript Projects
+- **Zero `.js`/`.jsx` files.** Everything is `.ts`/`.tsx`. If a file is JS, convert it.
+- **`tsconfig.json`: `strict: true`, `target: ES2022`, `module: NodeNext`.** No `any` without a justifying comment.
+- **Package manager: pnpm.** Never npm, never yarn. If a repo uses npm, migrate it.
+- **Formatter: prettier.** Linter: eslint with typescript-eslint. Both configured from day one.
+- **Test: vitest.** Not jest.
+- **Build: tsup** for libraries, **tsc** for applications.
+- **Import style: ESM only.** `"type": "module"` in package.json. No CommonJS (`require`, `module.exports`).
+
+### Python Projects
+- **Python 3.11 minimum.** Test on 3.11 and latest.
+- **Config: `pyproject.toml` only.** No setup.py, no setup.cfg, no requirements.txt. Dependencies go in `[project.dependencies]`.
+- **Build backend: hatch.** Not setuptools.
+- **Formatter: ruff format.** Linter: ruff. Not black, not flake8, not isort.
+- **Test: pytest** with `--tb=short`.
+- **Entry points: `[project.scripts]`** in pyproject.toml, not `if __name__ == "__main__"`.
+
+### Dart/Flutter Projects
+- **Dart only.** No embedded JS unless Flutter requires it.
+- **Package manager: pub.**
+- **Test: flutter_test.** Lint: `flutter analyze`.
+
+### Cross-Cutting Rules
+- **One language per project.** No Python+JS repos, no TS+JS mixing.
+- **One package manager per project.**
+- **Dependencies pinned to exact versions.**
+- **Editorconfig** in every repo root for consistent indentation.
+
+## Project Scaffolding
+
+When Kyan (operations agent) receives a project idea via Telegram:
+
+### Phase 1: Research
+1. Find latest official documentation for the relevant framework/library
+2. Check latest stable versions — pin all dependencies exactly
+3. Check GitHub Advisories for security issues in chosen dependencies
+4. Check community best practices (recent posts, discussions)
+5. Check if an existing KyaniteLabs repo solves a similar problem
+
+### Phase 2: Classify
+Determine project type and apply matching tech stack from the rules above.
+
+### Phase 3: Scaffold
+Create a private repo in Pastorsimon1798's personal account with:
+
+**Every project gets:**
+- `.gitignore`, `.editorconfig`, `LICENSE` (MIT)
+- `README.md` (what, why, run, test), `CHANGELOG.md`
+- `CLAUDE.md` + `AGENTS.md` (from org templates + project-specific)
+- `.cursorrules` + `.windsurfrules` (from org templates)
+- `.github/copilot-instructions.md`, `dependabot.yml`, `CODEOWNERS`
+- `.github/workflows/ci.yml` (lint + test + build, Blacksmith runner, concurrency, caching)
+- `.github/ISSUE_TEMPLATE/` (bug_report.yml, feature_request.yml, config.yml)
+
+**Python additionally:** pyproject.toml (hatch), src/project_name/__init__.py, tests/test_smoke.py
+**TypeScript additionally:** package.json (type:module), tsconfig.json (strict), eslint.config.js, .prettierrc, vitest.config.ts, src/index.ts
+
+### Phase 4: Verify
+- Initial commit + push
+- Branch protection on main
+- CI runs green
+- Parity check (CLAUDE.md = AGENTS.md)
+- Report to user: repo URL, stack, CI status
+
+## Local-First Inference (LM Studio)
+
+All KyaniteLabs projects that require an LLM must use local inference first. Server runs on Tailscale at `100.66.225.85:1234`.
+
+### Server Specs
+- **CPU**: AMD Ryzen AI Max 395 (Strix Halo) — 16 cores, 32 threads
+- **Memory**: 96GB unified RAM, 48GB VRAM (shared pool)
+
+### API Quick Reference
+| Action | Method | Endpoint | Body |
+|--------|--------|----------|------|
+| List loaded models | GET | `/v1/models` | — |
+| Load a model | POST | `/api/v1/models/load` | `{"model": "<key>"}` |
+| Unload a model | POST | `/api/v1/models/unload` | `{"instance_id": "<key>"}` |
+| Chat completion | POST | `/v1/chat/completions` | OpenAI-compatible |
+
+### Loading Rules
+1. **Check before loading** — call `GET /v1/models` first, reuse if already loaded
+2. **Do not touch models you did not load** — other agents or processes depend on them
+3. **Unload when done** — call unload endpoint after task completes
+4. **No downloading** — only use models already on the server
+5. **Always enable**: flash attention, KV cache quantization at Q8, full GPU offload
+6. **CPU thread pool: 10 threads** (formula: 16 cores × 0.625)
+7. **Memory budget**: keep 16GB free, max ~80GB for LLM usage
+
+### Inference Sampling Profiles
+| Profile | Temp | top_p | top_k | repeat_penalty | min_p |
+|---------|------|-------|-------|----------------|-------|
+| Coding | 0.0-0.1 | 0.95 | 40 | 1.0 | 0.05 |
+| Reasoning | 0.6-0.7 | 0.9 | 40 | 1.1 | 0.05 |
+| Creative | 0.8 | 0.92 | 45 | 1.05 | 0.04 |
+| Deterministic | 0.0 | 1.0 | 1 | 1.0 | 0.0 |