@dzhechkov/keysarium 1.7.19 → 1.7.20

package/templates/.claude/skills/feature-adr/references/agentic-qe/qcsd-production-swarm.md

@@ -1,108 +0,0 @@
-# QCSD Production Swarm (Condensed for Feature ADR)
-
-> Source: agentic-qe `qcsd-production-swarm` skill (Tier 3)
-> Condensed from 2781 lines — SAP/middleware-specific agents removed, universal patterns retained.
-
-## Purpose
-
-Post-release production health assessment and QCSD feedback loop closure.
-The only QCSD phase with dual responsibility: assessing current health AND
-closing the feedback loop back to earlier phases.
-
-## QCSD Phase Positioning
-
-| Phase | Swarm | Question | Decision |
-|-------|-------|----------|----------|
-| Ideation | qcsd-ideation-swarm | Should we build this? | GO / CONDITIONAL / NO-GO |
-| Refinement | qcsd-refinement-swarm | How should we test this? | READY / CONDITIONAL / NOT-READY |
-| Development | qcsd-development-swarm | Is code quality sufficient? | SHIP / CONDITIONAL / HOLD |
-| Verification | qcsd-cicd-swarm | Is this change safe to release? | RELEASE / REMEDIATE / BLOCK |
-| **Production** | **qcsd-production-swarm** | **Is release healthy?** | **HEALTHY / DEGRADED / CRITICAL** |
-
-## Core Agents (always run)
-
-| Agent | Domain | Role |
-|-------|--------|------|
-| qe-metrics-optimizer | learning-optimization | DORA metrics + SLA compliance |
-| qe-defect-predictor | defect-intelligence | Defect trend + density prediction |
-| qe-root-cause-analyzer | defect-intelligence | Incident RCA + severity assessment |
-
-## Conditional Agents (flag-based)
-
-| Flag | Agent | When to activate |
-|------|-------|------------------|
-| HAS_INFRASTRUCTURE_CHANGE | qe-chaos-engineer | Infra changes in release |
-| HAS_PERFORMANCE_SLA | qe-performance-tester | Performance SLAs defined |
-| HAS_REGRESSION_RISK | qe-regression-analyzer | Known regression risk areas |
-| HAS_RECURRING_INCIDENTS | qe-pattern-learner | Repeated incident patterns |
-
-## Feedback Agents (always run, SEQUENTIAL)
-
-| Order | Agent | Role |
-|-------|-------|------|
-| 1st | qe-learning-coordinator | Synthesize production findings |
-| 2nd | qe-transfer-specialist | Transfer learnings to Ideation + Refinement |
-
-## Decision Logic
-
-```
-CRITICAL (any condition):
-  - Active P0/P1 incidents
-  - DORA score < 0.4
-  - SLA compliance < 95%
-  - Accelerating defect trend (increasing + density > 5.0/KLOC)
-
-HEALTHY (all conditions):
-  - DORA score >= 0.7
-  - SLA compliance >= 99%
-  - No P0/P1/P2 incidents
-  - RCA completeness >= 80%
-  - Defect trend declining or stable
-  - Defect density <= 2.0/KLOC
-
-DEGRADED:
-  - Everything else (default)
-```
-
-## Execution Pattern (3 batches)
-
-```
-Batch 1 (parallel):  Core agents [3]
-  → WAIT for all to complete
-Batch 2 (parallel):  Conditional agents [0-4, based on flags]
-  → WAIT for all to complete
-Batch 3 (sequential): Feedback agents [2, strictly ordered]
-  → Learning coordinator THEN transfer specialist
-```
-
-## Enforcement Rules
-
-1. ALL THREE core agents MUST be spawned — no exceptions
-2. All parallel agents in a SINGLE message
-3. STOP and WAIT after each batch
-4. Conditional agents MUST spawn if flags are TRUE
-5. Decision logic applied EXACTLY as specified
-6. Both feedback agents MUST run SEQUENTIALLY
-7. Learning persistence MUST happen BEFORE feedback agents
-
-## Feedback Loop Closure
-
-Production swarm closes the QCSD loop by feeding findings back:
-
-```
-Production findings
-  → qe-learning-coordinator (synthesize)
-    → qe-transfer-specialist (distribute)
-      → Ideation phase (future GO/NO-GO informed by production data)
-      → Refinement phase (test strategy informed by production failures)
-```
-
-This is the key differentiator: no other QCSD phase feeds back to earlier phases.
-
-## Integration with Feature ADR
-
-When used in Direct Mode (full agentic-qe installed):
-- After Step 9 Fleet QE detects issues, production swarm feedback patterns
-  inform future `/feature-adr` runs via memory system
-- Historical production health data feeds Step 3.5 QCSD Ideation quality risks
-- Defect patterns from production inform Step 2 Research phase