NO-JIRA: AI-assisted test generation

.claude/STARTUP_CHECKLIST.md

@@ -0,0 +1,230 @@
+# Startup Checklist
+
+## CRITICAL ENFORCEMENT - NO EXCEPTIONS
+
+**ABSOLUTE REQUIREMENT: You MUST execute EVERY SINGLE STEP in agent YAML configurations. NO SKIPPING, NO COMBINING, NO SIMPLIFYING.**
+
+## EXECUTION EFFICIENCY REQUIREMENTS (MANDATORY)
+
+**CRITICAL: Must execute with maximum efficiency and minimal verbosity**
+
+**OUTPUT FORMAT ENFORCEMENT**:
+- Each step must have explicit output: `✅ Step X completed: [result]`
+- NEVER skip step numbers (Step 1 → Step 2 → Step 3 → Step 4)
+- Silent thinking execution (no verbose explanations), but clear step completion markers
+- NO process descriptions or step-by-step commentary
+
+**VIOLATION ENFORCEMENT**: If providing ANY explanatory text or verbose process descriptions, this is an EFFICIENCY VIOLATION and execution must restart with direct execution only.
+
+## FRESH START ENFORCEMENT
+
+**CRITICAL RULE: Every user request is a fresh execution, even if it happens in the SAME conversation.**
+
+### What MUST Be Re-read Each Request:
+- ✅ **MUST always re-read workflow_orchestrator.yaml** (user may have modified it)
+- ✅ **MUST always re-read the agent YAML config** (user may have modified it)
+- ✅ **MUST always re-read rule files if agent requires them** (rules may have changed)
+
+### What CAN Be Reused Across Requests:
+- ✅ **Generated artifacts** (test_artifacts/*) - unless user says "regenerate/re-create"
+- ✅ **E2E test code** (temp_repos/*) - unless user says "regenerate/re-create"
+- ✅ **Prerequisite validation** - orchestrator will check if artifacts exist
+
+### Key Principle:
+- **Configuration** = Always re-read (may have changed)
+- **Artifacts** = Reuse if exists (unless regenerate mode)
+- **Agent execution** = Follow orchestrator's prerequisite validation logic
+- Each user request = enforce mandatory startup protocol
+
+## MANDATORY Steps
+
+### Before Processing User Request - USE WORKFLOW ORCHESTRATOR
+- [ ] **MANDATORY: ALWAYS start by reading workflow_orchestrator.yaml**: Read `config/agents/workflow_orchestrator.yaml`
+- [ ] **Let orchestrator identify workflow automatically**: Use workflow definitions and trigger keywords
+- [ ] **Orchestrator will automatically**:
+  - Extract JIRA issue key from user request
+  - Match user request against workflow trigger keywords
+  - Detect REGENERATE mode ("re-create", "recreate", "re-generate", "regenerate")
+  - Validate prerequisites before execution
+  - Execute agents in correct sequence
+  - Generate workflow execution report
+
+### Manual Agent Identification (If Orchestrator Cannot Match)
+If workflow_orchestrator cannot identify a workflow, fallback to manual identification:
+- [ ] Identify request type and agent to execute:
+  - "Create test case for JIRA-XXX" → `test_case_generation` agent
+  - "Generate E2E code/test for JIRA-XXX" → `e2e_test_generation_openshift_private` agent
+  - "Run E2E tests for JIRA-XXX" → `test-executor` agent
+  - "Create E2E PR for JIRA-XXX" → `pr-submitter` agent
+
+### For ALL Agent Executions (Critical Rules)
+- [ ] **NEVER use Task tool for agent workflows**
+- [ ] **ALWAYS read agent YAML config directly using Read tool**
+- [ ] **Execute EACH step in agent config manually using available tools**
+- [ ] **Verify each step's output before proceeding to next step**
+- [ ] **Follow agent config instructions exactly - no skipping or simplifying**
+
+### CRITICAL: Rule Keyword Enforcement (MANDATORY)
+**When reading any rule files, configuration files, or guideline documents:**
+- [ ] **MANDATORY keywords = ABSOLUTE REQUIREMENT** - Must be executed without exception
+- [ ] **NEVER keywords = ABSOLUTE PROHIBITION** - Must never be violated under any circumstances
+- [ ] **CRITICAL keywords = STOP and verify** - Must pause and verify compliance before proceeding
+- [ ] **FORBIDDEN keywords = ABSOLUTE PROHIBITION** - Same as NEVER, must not be violated
+- [ ] If ANY of these keywords are violated → **STOP immediately and restart from beginning**
+
+**Enforcement Rules:**
+- When encountering "MANDATORY: Do X" → X must be done, no alternatives
+- When encountering "NEVER do Y" → Y is absolutely prohibited, no exceptions
+- When encountering "CRITICAL: Check Z" → Must verify Z before any further action
+- When encountering "FORBIDDEN: Action W" → Action W is absolutely prohibited
+- Violation of any keyword rule = **Complete execution failure, must restart**
+
+## STRICT EXECUTION PROTOCOL
+
+### Step-by-Step Execution Requirements
+1. **Read agent YAML config directly** - Use Read tool, never Task tool
+2. **Execute steps manually** - Follow config steps exactly, verify each step
+3. **Use correct directory structure** - test_artifacts/{COMPONENT}/{JIRA_KEY}/
+4. **Parallel execution mandatory** - Execute independent tool calls in parallel within same message
+
+### MANDATORY VERIFICATION FOR EACH STEP
+- [ ] Execute step exactly as written in YAML config
+- [ ] Verify step output/result
+- [ ] Log completion message: `✅ Step X completed: [brief result]`
+- [ ] Show step output/result (concise format)
+- [ ] Confirm step meets success criteria
+- [ ] Only then proceed to next step
+
+### VIOLATION DETECTION
+- [ ] If ANY step is skipped → STOP and restart from beginning
+- [ ] If steps are combined → STOP and restart from beginning  
+- [ ] If step verification is missing → STOP and restart from beginning
+- [ ] If different commands are used → STOP and restart from beginning
+- [ ] If verbose explanations provided → EFFICIENCY VIOLATION, restart
+
+## Execution Output Format
+
+### MANDATORY OUTPUT FORMAT FOR ALL AGENTS:
+```
+Step 1: Reading agent config
+✅ Step 1 completed: Agent config loaded from config/agents/<agent>.yaml
+
+Step 2: [Action from YAML step 1]
+✅ Step 2 completed: [Brief result]
+
+Step 3: [Action from YAML step 2]
+✅ Step 3 completed: [Brief result]
+
+[Continue for ALL steps in the YAML config...]
+```
+
+### EFFICIENCY RULES:
+- **60-70% fewer tokens than verbose mode**
+- **No process descriptions** - only step completions
+- **No explanatory text** - only execution results
+- **Brief result summaries** - 1-2 sentences maximum per step
+- **Clear step numbers** - never skip sequence
+
+## Agent Execution Examples (Using Workflow Orchestrator)
+
+### Example 1: Complete Flow (Recommended Approach)
+```
+User: "Generate test cases and run them for HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+✅ Step 1 completed: Orchestrator config loaded
+
+Step 2: Parse user request
+✅ Step 2 completed: JIRA key=HIVE-2883, Workflow=full_flow, Regenerate=false
+
+Step 3: Validate prerequisites
+✅ Step 3 completed: All prerequisites validated
+
+Step 4: Execute workflow agents
+  4.1: Executing test_case_generation agent
+       ✅ Agent completed: 4 files generated
+  4.2: Executing e2e_test_generation_openshift_private agent
+       ✅ Agent completed: E2E code integrated
+  4.3: Executing test-executor agent
+       ✅ Agent completed: Tests executed
+
+Step 5: Generate workflow report
+✅ Workflow completed: full_flow for HIVE-2883 (3 agents executed)
+```
+
+### Example 2: Single Agent via Orchestrator
+```
+User: "Create test case for HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+✅ Step 1 completed: Orchestrator config loaded
+
+Step 2: Parse user request
+✅ Step 2 completed: Workflow identified=test_case_only
+
+Step 3: Execute workflow
+✅ Agent test_case_generation completed
+
+Output: 4 files in test_artifacts/hive/HIVE-2883/
+```
+
+### Example 3: Regenerate Mode via Orchestrator
+```
+User: "re-create e2e test for HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+✅ Step 1 completed: Orchestrator config loaded
+
+Step 2: Parse user request
+✅ Step 2 completed: REGENERATE mode detected, Workflow=e2e_generation
+
+Step 3: Validate prerequisites
+⚠️ REGENERATE mode - skipping prerequisite checks
+
+Step 4: Execute workflow
+✅ Agent e2e_test_generation_openshift_private completed (overwrite mode)
+```
+
+### Example 4: E2E Test Generation (Shows Sub-Agent Orchestration)
+```
+User: "Generate E2E code for HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+✅ Workflow identified: e2e_generation
+
+Step 2: Execute e2e_test_generation_openshift_private agent
+  → This agent is also an orchestrator, executing 4 sub-agents:
+  2.1: e2e_validation_agent
+  2.2: repository_setup_agent  
+  2.3: e2e_code_generation_agent
+  2.4: e2e_quality_check_agent
+✅ E2E generation workflow completed
+```
+
+### Example 5: PR Submission via Orchestrator
+```
+User: "Create PR for HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+✅ Workflow identified: pr_submission
+
+Step 2: Validate prerequisites
+✅ E2E code exists in openshift-tests-private
+
+Step 3: Execute pr-submitter agent
+✅ PR created and submitted
+
+Output: GitHub PR link
+```
+
+### Example 6: Manual Fallback (Orchestrator Cannot Match)
+```
+User: "Do something custom with HIVE-2883"
+
+Step 1: Read workflow_orchestrator.yaml
+⚠️ No workflow matched - falling back to manual identification
+
+Step 2: Manual agent identification
+→ AI determines appropriate agent based on context
+→ Execute selected agent directly
+```