Feat/l6 auto evolution

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,55 @@
+---
+name: Bug report 🐛
+about: Report a defect in the factory line (pipeline, prompts, CI)
+title: "[BUG] "
+labels: bug
+assignees: ""
+---
+
+## Summary
+
+A clear, one-paragraph description of the bug.
+
+## Where did it fail?
+
+- **Mode:** full / express
+- **Phase:** 0 / 1 / 2 / 3 / 4 / 5 / 6 / 7
+- **Run ID:** `run-YYYYMMDD-HHMMSS` (from the factory banner)
+- **Checkpoint choice (if relevant):** approve / modify / skip-all / abort
+
+## Steps to reproduce
+
+1. Run: `dark factory "..."` (paste the exact goal)
+2. Approve checkpoints: (list what you clicked)
+3. Observe: (what happened)
+
+## Expected behavior
+
+What should have happened instead?
+
+## Actual behavior
+
+What happened? Include the most relevant output.
+
+```text
+(paste output here)
+```
+
+## Sealed-envelope integrity check (required)
+
+- [ ] The Lead Engineer never saw sealed test code
+- [ ] The user was never shown sealed test code before validation
+- [ ] Hardening included only failure messages (no test source)
+
+If any of the above are false, treat this as a potential security issue and also follow **SECURITY.md**.
+
+## Environment
+
+- OS:
+- GitHub Copilot CLI version:
+- Dark Factory version: 0.1.0
+
+## Additional context
+
+- Relevant `config.yml` changes (if any):
+- Repo stack (Node/Python/Go/etc):

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,39 @@
+---
+name: Feature request 💡
+about: Propose an improvement to the factory line
+title: "[FEAT] "
+labels: enhancement
+assignees: ""
+---
+
+## Problem
+
+What problem are you trying to solve? Who is affected?
+
+## Proposal
+
+Describe the change you want. Include a concrete example of the desired behavior.
+
+## Acceptance criteria
+
+- [ ] Criterion 1
+- [ ] Criterion 2
+
+## Pipeline impact
+
+Which part of the line changes?
+
+- [ ] Orchestrator (`SKILL.md`)
+- [ ] Agent prompt (`agents/*.md`)
+- [ ] Templates (`templates/*.md`)
+- [ ] Protocols (`protocols/*.md`)
+- [ ] Config (`config.yml`)
+- [ ] CI (`.github/workflows/validate.yml`)
+
+## Safety / sealed-envelope considerations
+
+Explain why this feature does not compromise sealed-envelope isolation.
+
+## Alternatives considered
+
+List alternatives and why you did not choose them.

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,41 @@
+## Summary
+
+What does this PR change? Link issues.
+
+Fixes #
+
+## Type
+
+- [ ] Bug fix
+- [ ] New feature
+- [ ] Documentation
+- [ ] CI / Infra
+
+## Factory impact
+
+- Affects phases: 0 / 1 / 2 / 3 / 4 / 5 / 6 / 7
+- Agent prompts touched (if any):
+
+## QA checklist (from docs/TESTING.md)
+
+- [ ] 🏭 Factory banner displays on start (run ID, mode, agent lineup)
+- [ ] 📋 PRD.md produced with user stories and acceptance criteria
+- [ ] 🏗️ ARCH.md produced with components, file structure, tech choices
+- [ ] 🔒 Sealed tests generated from PRD only (never from code/arch)
+- [ ] 🔒 SHA-256 hash of sealed directory displayed at Phase 2
+- [ ] 👩‍💻 Engineer never accesses `.factory/sealed/` during build
+- [ ] ✅ Both test suites run in Phase 4
+- [ ] 📊 GAP-REPORT.md produced with gap score
+- [ ] 🔧 Hardening sends failure messages only (no test code)
+- [ ] 🔧 Hardening caps at configured max cycles, then escalates
+- [ ] 📋 Checkpoints use `ask_user` with exactly 4 choices
+- [ ] ⏭️ Skip-all still shows final delivery checkpoint
+- [ ] 💾 state.json written on every phase transition
+- [ ] 🔄 Resume works from state.json
+- [ ] 🛑 Abort cleans up worktree and branch
+- [ ] ⚡ Express mode skips PRD/arch, one checkpoint at delivery
+- [ ] 🗄️ SQL tables track runs and phase results
+
+## Notes for reviewers
+
+Anything risky, subtle, or worth double-checking?

.github/copilot-instructions.md

@@ -1,28 +1,39 @@
 # Copilot Instructions — Dark Factory
 
-This repository contains the **Dark Factory** Copilot CLI skill — an agentic build system that orchestrates specialist AI agents through a sealed-envelope testing pipeline.
+This repository contains **Dark Factory**, a GitHub Copilot CLI skill that orchestrates an agent pipeline with **sealed-envelope testing**.
 
-## Key Concepts
+## File map
 
-- **Sealed-envelope testing**: QA agent generates tests from the PRD *before* code is written. Building agents never see these tests. The gap between builder tests and sealed tests measures build quality.
-- **Checkpoint-gated pipeline**: Each phase runs autonomously, pausing for human approval at phase boundaries.
-- **Git worktree isolation**: All building happens in an isolated worktree. Nothing touches the user's working directory until they approve.
+| File/Dir | Purpose |
+|---|---|
+| `SKILL.md` | Factory Manager orchestrator prompt (the brain) |
+| `agents/*.md` | Specialist agent prompts |
+| `templates/*.md` | Artifact output formats |
+| `protocols/*.md` | Protocol invariants (sealed envelope, checkpoints) |
+| `config.yml` | Tunables (models, thresholds, timeouts) |
+| `catalog.yml` | Skill metadata + file references |
+| `docs/TESTING.md` | Playbooks + QA checklist |
+| `docs/ADR.md` | Why these decisions |
 
-## File Map
+## Non-negotiables
 
-| File | Purpose |
-|------|---------|
-| `SKILL.md` | Factory Manager orchestrator prompt (the brain) |
-| `catalog.yml` | Skill metadata for Copilot CLI registration |
-| `config.yml` | Tunable settings (models, thresholds, checkpoints) |
-| `agents/*.md` | Specialist agent prompts (product, arch, qa, eng) |
-| `templates/*.md` | Structured output templates for agent artifacts |
-| `protocols/*.md` | Reusable protocol definitions |
-
-## Rules
-
-1. **Never modify sealed test isolation** — the sealed directory must remain invisible to building agents
-2. **Agents are stateless** — each `task()` call gets a clean context. Pass EXACTLY what the agent needs via the handoff manifest.
-3. **State must survive crashes** — every phase transition writes to `state.json`
-4. **Express mode must work** — quick tasks should complete in ~60 seconds with one checkpoint
-5. **Keep agent prompts under 200 lines** — focused, not bloated
+1. **Sealed envelope is sacred.** The Lead Engineer must never see sealed tests; the QA Sealed agent must never see code.
+2. **Config is the source of truth.** Never hardcode model names or tunables inside prompts.
+3. **Agent prompts <= 200 lines.** If a prompt grows, split responsibilities instead.
+4. **Just a skill.** Do not add runtime code, package managers, telemetry, dashboards, or plugin systems.
+5. **Worktree isolation.** All build work happens under `.factory/` until delivery approval.
+
+## Prohibited actions
+
+- Exposing sealed test contents to the user before Phase 4.
+- Passing sealed test source code into any builder/hardening prompt.
+- Editing user files outside the factory worktree.
+
+## PR requirements
+
+Before opening a PR:
+
+- Run Playbook 1 (Full) and Playbook 2 (Express) from `docs/TESTING.md`.
+- Ensure `catalog.yml` references are valid.
+- Ensure YAML parses (`config.yml`, `catalog.yml`).
+- Ensure CI validate workflow passes.

.github/workflows/validate.yml

@@ -0,0 +1,129 @@
+name: Validate Dark Factory
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.x'
+
+      - name: Install YAML parser
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install pyyaml
+
+      - name: Validate YAML syntax
+        run: |
+          python - <<'PY'
+          import yaml
+          for path in ("config.yml", "catalog.yml"):
+              with open(path, "r", encoding="utf-8") as f:
+                  yaml.safe_load(f)
+          print("✅ YAML valid")
+          PY
+
+      - name: Verify catalog references exist
+        run: |
+          python - <<'PY'
+          import pathlib, sys
+          import yaml
+
+          root = pathlib.Path(".")
+          catalog = yaml.safe_load(root.joinpath("catalog.yml").read_text(encoding="utf-8"))
+          links = catalog.get("links", {}) or {}
+
+          referenced = []
+          if links.get("skill_file"):
+              referenced.append(links["skill_file"])
+          referenced.extend(list(links.get("agents", []) or []))
+
+          missing = [p for p in referenced if not root.joinpath(p).is_file()]
+          if missing:
+              print("::error::catalog.yml references missing files")
+              for p in missing:
+                  print(f"- {p}")
+              sys.exit(1)
+
+          print("✅ Catalog references valid")
+          PY
+
+      - name: Check agent line counts (<= 200)
+        run: |
+          for file in agents/*.md; do
+            lines=$(wc -l < "$file")
+            if [ "$lines" -gt 200 ]; then
+              echo "::error file=$file::Agent prompt exceeds 200 lines ($lines lines)"
+              exit 1
+            fi
+          done
+          echo "✅ Agent prompts within limit"
+
+      - name: Verify SKILL.md frontmatter
+        run: |
+          python - <<'PY'
+          import pathlib, sys
+
+          lines = pathlib.Path("SKILL.md").read_text(encoding="utf-8").splitlines()
+          if not lines or lines[0].strip() != "---":
+              print("::error file=SKILL.md::Missing frontmatter start (---)")
+              sys.exit(1)
+
+          try:
+              end = lines[1:200].index("---") + 1
+          except ValueError:
+              print("::error file=SKILL.md::Missing frontmatter end (---)")
+              sys.exit(1)
+
+          if end < 3:
+              print("::error file=SKILL.md::Frontmatter too short")
+              sys.exit(1)
+
+          print("✅ Frontmatter present")
+          PY
+
+      - name: Set up Node (markdown lint)
+        uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+
+      - name: Install markdownlint
+        run: npm install -g markdownlint-cli2
+
+      - name: Markdown lint
+        run: |
+          cat > .markdownlint-cli2.jsonc <<'JSON'
+          {
+            "config": {
+              "default": true,
+              "MD013": false,
+              "MD022": false,
+              "MD024": false,
+              "MD025": false,
+              "MD029": false,
+              "MD031": false,
+              "MD032": false,
+              "MD033": false,
+              "MD036": false,
+              "MD040": false,
+              "MD041": false,
+              "MD056": false,
+              "MD058": false,
+              "MD060": false
+            },
+            "globs": ["**/*.md"],
+            "ignores": ["**/.factory/**", "**/node_modules/**"]
+          }
+          JSON
+          markdownlint-cli2

.gitignore

@@ -0,0 +1,28 @@
+# Dark Factory runtime artifacts
+.factory/
+
+# Environment
+.env
+.env.local
+
+# Python
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.Python
+env/
+venv/
+
+# Node
+node_modules/
+npm-debug.log
+yarn-error.log
+
+# OS / IDE
+.DS_Store
+Thumbs.db
+.vscode/
+.idea/
+*.swp
+*.swo

AGENTS.md

@@ -6,15 +6,16 @@ This file tells any AI agent how to work effectively on the Dark Factory codebas
 
 ## Architecture
 
-Dark Factory is a **Copilot CLI skill** that orchestrates 5 specialist agents through a checkpoint-gated pipeline with sealed-envelope testing.
+Dark Factory is a **Copilot CLI skill** that orchestrates 6 specialist agents through a checkpoint-gated pipeline with sealed-envelope testing.
 
 ```
 SKILL.md (Factory Manager / Orchestrator)
-  ├── agents/product-mgr.md   → Phase 1: writes PRD.md
-  ├── agents/architect.md     → Phase 2b: writes ARCH.md
-  ├── agents/qa-sealed.md     → Phase 2a: writes sealed tests (parallel with arch)
-  ├── agents/lead-eng.md      → Phase 3: writes code + open tests
-  └── agents/qa-validator.md  → Phase 4: runs sealed tests, gap analysis
+  ├── agents/product-mgr.md       → Phase 1: writes PRD.md
+  ├── agents/architect.md         → Phase 2b: writes ARCH.md
+  ├── agents/qa-sealed.md         → Phase 2a: writes sealed tests (parallel with arch)
+  ├── agents/lead-eng.md          → Phase 3: writes code + open tests
+  ├── agents/qa-validator.md      → Phase 4: runs sealed tests, gap analysis
+  └── agents/outcome-evaluator.md → Phase 7: evaluates post-ship outcomes
 ```
 
 Each agent is called via `task()` with explicit inputs. Agents are stateless — they receive ONLY what the handoff manifest specifies.
@@ -27,6 +28,7 @@ Each agent is called via `task()` with explicit inputs. Agents are stateless —
 |----------|-------|-------------|
 | `SKILL.md` | Orchestrator logic | Most critical file. Changes affect all phases. |
 | `agents/*.md` | Individual agent prompts | Change one agent at a time. Test with a factory run. |
+| `agents/outcome-evaluator.md` | Post-ship analysis prompt | Keep KPI math aligned with PRD template + config `outcome_evaluation` settings. |
 | `config.yml` | User-tunable settings | Never hardcode values that belong here. |
 | `templates/*.md` | Output format templates | Keep in sync with agent prompts that reference them. |
 | `protocols/*.md` | Reusable protocol defs | Referenced by SKILL.md and agent prompts. |