Add agentic workflow layer with composable phases (#4159)

## Summary

- Adds the `agentic/` layer with composable workflow phases: plan,
build, test, review, document, and ship
- Each phase is a standalone script (`uv run
agentic/workflows/<phase>.py`) that can run independently or be piped
together
- Orchestrators chain phases into pipelines (`plan_build.py`,
`plan_build_test.py`, ..., `plan_build_test_review_document_ship.py`)
- Shared modules (`agent.py`, `state.py`, `template.py`,
`orchestrator.py`) handle execution, state persistence, and template
rendering
- Supports multiple CLI backends (claude, copilot, gemini) with
abstracted model tiers (small/medium/large)

## Test plan

- [x] All 1012 existing tests pass (`uv run python -m pytest tests/ -x`)
- [x] New files compile cleanly (`py_compile`)
- [x] No stale references to removed files
- [ ] Manual: run `uv run agentic/workflows/ship.py --run-id <id>` to
verify commit + PR flow
- [ ] Manual: run `uv run
agentic/workflows/plan_build_test_review_document_ship.py "<prompt>"`
end-to-end

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: MarkusNeusinger

.claude/commands/bug.md

@@ -6,12 +6,16 @@
       "Bash",
       "WebFetch",
       "mcp__plugin_serena_serena__*",
-      "mcp__plugin_context7_context7__*"
+      "mcp__plugin_context7_context7__*",
+      "mcp__plugin_playwright_playwright__*"
     ],
     "ask": [
       "Bash(git commit *)",
       "Bash(git push *)"
     ],
     "deny": []
+  },
+  "env": {
+    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
   }
 }

.claude/commands/check.md

@@ -41,3 +41,24 @@ PORT=8000
 
 # Anthropic API key for Claude
 # ANTHROPIC_API_KEY=sk-ant-...
+
+# ============================================================================
+# CLI Model Tier Configuration (optional)
+# ============================================================================
+# Override the default model for each tier per CLI tool.
+# Tiers: small (fast/cheap), medium (balanced), large (complex)
+
+# Claude Code models
+# CLI_MODEL_CLAUDE_SMALL=haiku
+# CLI_MODEL_CLAUDE_MEDIUM=sonnet
+# CLI_MODEL_CLAUDE_LARGE=opus
+
+# GitHub Copilot models
+# CLI_MODEL_COPILOT_SMALL=gpt-4o-mini
+# CLI_MODEL_COPILOT_MEDIUM=gpt-4o
+# CLI_MODEL_COPILOT_LARGE=o1
+
+# Gemini CLI models
+# CLI_MODEL_GEMINI_SMALL=gemini-2.0-flash
+# CLI_MODEL_GEMINI_MEDIUM=gemini-2.0-flash-thinking
+# CLI_MODEL_GEMINI_LARGE=gemini-2.5-pro

.claude/commands/chore.md

@@ -112,7 +112,7 @@ jobs:
       run: |
         uv run pytest tests/unit tests/integration tests/e2e \
           -v --tb=short \
-          --cov=core --cov=api --cov=automation \
+          --cov=core --cov=api --cov=automation --cov=agentic/workflows/modules \
           --cov-report=term-missing \
           --cov-report=xml
       # Tests include:

.claude/commands/feature.md

@@ -227,6 +227,9 @@ nul
 # Project-specific
 temp/
 
+# Agentic runs output
+agentic/runs/
+
 # Additional environment files
 .env.local
 .env.production