QCSD Production Swarm v1.0
Post-release production health assessment and QCSD feedback loop closure.
Overview
The Production Swarm assesses release health in the live production environment using
DORA metrics, incident RCA, defect prediction, and cross-phase feedback loops. It renders
a HEALTHY / DEGRADED / CRITICAL decision and is the only QCSD phase with dual
responsibility: assessing current production health AND closing the feedback loop
back to Ideation and Refinement phases.
QCSD Phase Positioning
| Phase |
Swarm |
Decision |
When |
| Ideation |
qcsd-ideation-swarm |
GO / CONDITIONAL / NO-GO |
PI/Sprint Planning |
| Refinement |
qcsd-refinement-swarm |
READY / CONDITIONAL / NOT-READY |
Sprint Refinement |
| Development |
qcsd-development-swarm |
SHIP / CONDITIONAL / HOLD |
During Sprint |
| Verification |
qcsd-cicd-swarm |
RELEASE / REMEDIATE / BLOCK |
Pre-Release / CI-CD |
| Production |
qcsd-production-swarm |
HEALTHY / DEGRADED / CRITICAL |
Post-Release |
Parameters
TELEMETRY_DATA: Path to production telemetry, incident reports, and DORA metrics (required)RELEASE_ID: Release identifier for tracking (optional)OUTPUT_FOLDER: Where to save reports (default: ${PROJECT_ROOT}/Agentic QCSD/production/)SLA_DEFINITIONS: Path to SLA/SLO target definitions (optional)
ENFORCEMENT RULES - READ FIRST
| Rule |
Enforcement |
| E1 |
You MUST spawn ALL THREE core agents in Step 2. No exceptions. |
| E2 |
You MUST put all parallel Task calls in a SINGLE message. |
| E3 |
You MUST STOP and WAIT after each batch. No proceeding early. |
| E4 |
You MUST spawn conditional agents if flags are TRUE. No skipping. |
| E5 |
You MUST apply HEALTHY/DEGRADED/CRITICAL logic exactly as specified in Step 5. |
| E6 |
You MUST generate the full report structure. No abbreviated versions. |
| E7 |
Each agent MUST read its reference files before analysis. |
| E8 |
You MUST run BOTH feedback agents in Step 8 SEQUENTIALLY. Always. Both agents. |
| E9 |
You MUST execute Step 7 learning persistence. No skipping. |
PROHIBITED BEHAVIORS:
- Summarizing instead of spawning agents
- Skipping agents "for brevity"
- Proceeding before background tasks complete
- Providing your own analysis instead of spawning specialists
- Omitting report sections or using placeholder text
Step Execution Protocol
This skill uses a micro-file step architecture. Each step is a self-contained file
loaded one at a time to avoid "lost in the middle" context degradation.
Execute steps sequentially by reading each step file with the Read tool.
Steps
- Flag Detection --
steps/01-flag-detection.md -- Retrieve CI/CD signals, detect telemetry source, evaluate all 7 flags
- Core Agents --
steps/02-core-agents.md -- Spawn qe-metrics-optimizer, qe-defect-predictor, qe-root-cause-analyzer in parallel
- Batch 1 Results --
steps/03-batch1-results.md -- Wait for core agents, extract all metrics
- Conditional Agents --
steps/04-conditional-agents.md -- Spawn flagged conditional agents in parallel
- Decision Synthesis --
steps/05-decision-synthesis.md -- Apply HEALTHY/DEGRADED/CRITICAL logic
- Report Generation --
steps/06-report-generation.md -- Generate executive summary and full report
- Learning Persistence --
steps/07-learning-persistence.md -- Store findings to memory, save persistence record
- Feedback Loop --
steps/08-feedback-loop.md -- Run learning coordinator then transfer specialist (sequential)
- Final Output --
steps/09-final-output.md -- Display completion summary with all scores
Execution Instructions
- Use the Read tool to load the current step file (e.g.,
Read({ file_path: ".claude/skills/qcsd-production-swarm/steps/01-flag-detection.md" }))
- Execute the step's instructions completely
- Verify all success criteria are met before proceeding
- Pass the step's output as context to the next step
- If a step fails, halt and report the failure point -- do not skip ahead
Resume Support
To resume from a specific step: specify --from-step N and the orchestrator will
skip to step N. Ensure you have the required prerequisite data from prior steps.
Agent Inventory
| Agent |
Type |
Domain |
Batch |
| qe-metrics-optimizer |
Core (always) |
learning-optimization |
1 |
| qe-defect-predictor |
Core (always) |
defect-intelligence |
1 |
| qe-root-cause-analyzer |
Core (always) |
defect-intelligence |
1 |
| qe-chaos-engineer |
Conditional (HAS_INFRASTRUCTURE_CHANGE) |
chaos-resilience |
2 |
| qe-performance-tester |
Conditional (HAS_PERFORMANCE_SLA) |
chaos-resilience |
2 |
| qe-regression-analyzer |
Conditional (HAS_REGRESSION_RISK) |
defect-intelligence |
2 |
| qe-pattern-learner |
Conditional (HAS_RECURRING_INCIDENTS) |
defect-intelligence |
2 |
| qe-middleware-validator |
Conditional (HAS_MIDDLEWARE) |
enterprise-integration |
2 |
| qe-sap-rfc-tester |
Conditional (HAS_SAP_INTEGRATION) |
enterprise-integration |
2 |
| qe-sod-analyzer |
Conditional (HAS_AUTHORIZATION) |
enterprise-integration |
2 |
| qe-learning-coordinator |
Feedback (always, sequential) |
learning-optimization |
3 |
| qe-transfer-specialist |
Feedback (always, sequential) |
learning-optimization |
3 |
Total: 12 agents (3 core + 7 conditional + 2 feedback)
Quality Gate Thresholds
| Metric |
HEALTHY |
DEGRADED |
CRITICAL |
| DORA Score |
>= 0.7 |
0.4 - 0.69 |
< 0.4 |
| SLA Compliance |
>= 99% |
95 - 98.9% |
< 95% |
| Incident Severity |
P3/P4/NONE |
P2 |
P0/P1 |
| Defect Trend |
declining/stable |
stable (density > 2) |
increasing + density > 5 |
| RCA Completeness |
>= 80% |
50 - 79% |
< 50% |
Report Filename Mapping
| Agent |
Report Filename |
Step |
| qe-metrics-optimizer |
02-dora-metrics.md |
2 |
| qe-defect-predictor |
03-defect-prediction.md |
2 |
| qe-root-cause-analyzer |
04-root-cause-analysis.md |
2 |
| qe-chaos-engineer |
05-chaos-resilience.md |
4 |
| qe-performance-tester |
06-performance-sla.md |
4 |
| qe-regression-analyzer |
07-regression-analysis.md |
4 |
| qe-pattern-learner |
08-pattern-analysis.md |
4 |
| Learning Persistence |
09-learning-persistence.json |
7 |
| qe-middleware-validator |
10-middleware-health.md |
4 |
| qe-sap-rfc-tester |
11-sap-health.md |
4 |
| qe-sod-analyzer |
12-sod-compliance.md |
4 |
| Feedback agents |
13-feedback-loops.md |
8 |
| Synthesis |
01-executive-summary.md |
6 |
Execution Model Options
| Model |
When to Use |
Agent Spawn |
| Task Tool (PRIMARY) |
Claude Code sessions |
Task({ subagent_type, run_in_background: true }) |
| MCP Tools |
MCP server available |
fleet_init({}) / task_submit({}) |
| CLI |
Terminal/scripts |
swarm init / agent spawn |
Key Principle
Production health is measured by outcomes, not intentions. This swarm provides
evidence-based production assessment and closes the QCSD feedback loop.
1---
2name: qcsd-production-swarm
3description: "Use when assessing post-release production health with DORA metrics, root cause analysis, defect prediction, or cross-phase feedback loops in the QCSD Production phase."
4category: qcsd-phases
5priority: critical
6version: 1.0.0
7tokenEstimate: 32000
8# DDD Domain Mapping (from QCSD-AGENTIC-QE-MAPPING-FRAMEWORK.md)
9domains:
10 primary:
11 - domain: learning-optimization
12 agents: [qe-metrics-optimizer]
13 - domain: defect-intelligence
14 agents: [qe-defect-predictor, qe-root-cause-analyzer]
15 conditional:
16 - domain: chaos-resilience
17 agents: [qe-chaos-engineer, qe-performance-tester]
18 - domain: defect-intelligence
19 agents: [qe-regression-analyzer, qe-pattern-learner]
20 - domain: enterprise-integration
21 agents: [qe-middleware-validator, qe-sap-rfc-tester, qe-sod-analyzer]
22 feedback:
23 - domain: learning-optimization
24 agents: [qe-learning-coordinator, qe-transfer-specialist]
25# Agent Inventory
26agents:
27 core: [qe-metrics-optimizer, qe-defect-predictor, qe-root-cause-analyzer]
28 conditional: [qe-chaos-engineer, qe-performance-tester, qe-regression-analyzer, qe-pattern-learner, qe-middleware-validator, qe-sap-rfc-tester, qe-sod-analyzer]
29 feedback: [qe-learning-coordinator, qe-transfer-specialist]
30 total: 12
31 sub_agents: 0
32skills: [shift-right-testing, chaos-engineering-resilience, quality-metrics, performance-testing, holistic-testing-pact]
33# Execution Models (Task Tool is PRIMARY)
34execution:
35 primary: task-tool
36 alternatives: [mcp-tools, cli]
37swarm_pattern: true
38parallel_batches: 3
39last_updated: 2026-02-17
40enforcement_level: strict
41tags: [qcsd, production, telemetry, dora, rca, defect-prediction, feedback-loop, learning, swarm, parallel, ddd]
42trust_tier: 3
43validation:
44 schema_path: schemas/output.json
45 validator_path: scripts/validate-config.json
46 eval_path: evals/qcsd-production-swarm.yaml
47---
48
49# QCSD Production Swarm v1.0
50
51Post-release production health assessment and QCSD feedback loop closure.
52
53---
54
55## Overview
56
57The Production Swarm assesses release health in the live production environment using
58DORA metrics, incident RCA, defect prediction, and cross-phase feedback loops. It renders
59a HEALTHY / DEGRADED / CRITICAL decision and is the only QCSD phase with dual
60responsibility: assessing current production health AND closing the feedback loop
61back to Ideation and Refinement phases.
62
63### QCSD Phase Positioning
64
65| Phase | Swarm | Decision | When |
66|-------|-------|----------|------|
67| Ideation | qcsd-ideation-swarm | GO / CONDITIONAL / NO-GO | PI/Sprint Planning |
68| Refinement | qcsd-refinement-swarm | READY / CONDITIONAL / NOT-READY | Sprint Refinement |
69| Development | qcsd-development-swarm | SHIP / CONDITIONAL / HOLD | During Sprint |
70| Verification | qcsd-cicd-swarm | RELEASE / REMEDIATE / BLOCK | Pre-Release / CI-CD |
71| **Production** | **qcsd-production-swarm** | **HEALTHY / DEGRADED / CRITICAL** | **Post-Release** |
72
73### Parameters
74
75- `TELEMETRY_DATA`: Path to production telemetry, incident reports, and DORA metrics (required)
76- `RELEASE_ID`: Release identifier for tracking (optional)
77- `OUTPUT_FOLDER`: Where to save reports (default: `${PROJECT_ROOT}/Agentic QCSD/production/`)
78- `SLA_DEFINITIONS`: Path to SLA/SLO target definitions (optional)
79
80---
81
82## ENFORCEMENT RULES - READ FIRST
83
84| Rule | Enforcement |
85|------|-------------|
86| **E1** | You MUST spawn ALL THREE core agents in Step 2. No exceptions. |
87| **E2** | You MUST put all parallel Task calls in a SINGLE message. |
88| **E3** | You MUST STOP and WAIT after each batch. No proceeding early. |
89| **E4** | You MUST spawn conditional agents if flags are TRUE. No skipping. |
90| **E5** | You MUST apply HEALTHY/DEGRADED/CRITICAL logic exactly as specified in Step 5. |
91| **E6** | You MUST generate the full report structure. No abbreviated versions. |
92| **E7** | Each agent MUST read its reference files before analysis. |
93| **E8** | You MUST run BOTH feedback agents in Step 8 SEQUENTIALLY. Always. Both agents. |
94| **E9** | You MUST execute Step 7 learning persistence. No skipping. |
95
96**PROHIBITED BEHAVIORS:**
97- Summarizing instead of spawning agents
98- Skipping agents "for brevity"
99- Proceeding before background tasks complete
100- Providing your own analysis instead of spawning specialists
101- Omitting report sections or using placeholder text
102
103---
104
105## Step Execution Protocol
106
107This skill uses a micro-file step architecture. Each step is a self-contained file
108loaded one at a time to avoid "lost in the middle" context degradation.
109
110**Execute steps sequentially by reading each step file with the Read tool.**
111
112### Steps
113
1141. **Flag Detection** -- `steps/01-flag-detection.md` -- Retrieve CI/CD signals, detect telemetry source, evaluate all 7 flags
1152. **Core Agents** -- `steps/02-core-agents.md` -- Spawn qe-metrics-optimizer, qe-defect-predictor, qe-root-cause-analyzer in parallel
1163. **Batch 1 Results** -- `steps/03-batch1-results.md` -- Wait for core agents, extract all metrics
1174. **Conditional Agents** -- `steps/04-conditional-agents.md` -- Spawn flagged conditional agents in parallel
1185. **Decision Synthesis** -- `steps/05-decision-synthesis.md` -- Apply HEALTHY/DEGRADED/CRITICAL logic
1196. **Report Generation** -- `steps/06-report-generation.md` -- Generate executive summary and full report
1207. **Learning Persistence** -- `steps/07-learning-persistence.md` -- Store findings to memory, save persistence record
1218. **Feedback Loop** -- `steps/08-feedback-loop.md` -- Run learning coordinator then transfer specialist (sequential)
1229. **Final Output** -- `steps/09-final-output.md` -- Display completion summary with all scores
123
124### Execution Instructions
125
1261. Use the Read tool to load the current step file (e.g., `Read({ file_path: ".claude/skills/qcsd-production-swarm/steps/01-flag-detection.md" })`)
1272. Execute the step's instructions completely
1283. Verify all success criteria are met before proceeding
1294. Pass the step's output as context to the next step
1305. If a step fails, halt and report the failure point -- do not skip ahead
131
132### Resume Support
133
134To resume from a specific step: specify `--from-step N` and the orchestrator will
135skip to step N. Ensure you have the required prerequisite data from prior steps.
136
137---
138
139## Agent Inventory
140
141| Agent | Type | Domain | Batch |
142|-------|------|--------|-------|
143| qe-metrics-optimizer | Core (always) | learning-optimization | 1 |
144| qe-defect-predictor | Core (always) | defect-intelligence | 1 |
145| qe-root-cause-analyzer | Core (always) | defect-intelligence | 1 |
146| qe-chaos-engineer | Conditional (HAS_INFRASTRUCTURE_CHANGE) | chaos-resilience | 2 |
147| qe-performance-tester | Conditional (HAS_PERFORMANCE_SLA) | chaos-resilience | 2 |
148| qe-regression-analyzer | Conditional (HAS_REGRESSION_RISK) | defect-intelligence | 2 |
149| qe-pattern-learner | Conditional (HAS_RECURRING_INCIDENTS) | defect-intelligence | 2 |
150| qe-middleware-validator | Conditional (HAS_MIDDLEWARE) | enterprise-integration | 2 |
151| qe-sap-rfc-tester | Conditional (HAS_SAP_INTEGRATION) | enterprise-integration | 2 |
152| qe-sod-analyzer | Conditional (HAS_AUTHORIZATION) | enterprise-integration | 2 |
153| qe-learning-coordinator | Feedback (always, sequential) | learning-optimization | 3 |
154| qe-transfer-specialist | Feedback (always, sequential) | learning-optimization | 3 |
155
156**Total: 12 agents (3 core + 7 conditional + 2 feedback)**
157
158---
159
160## Quality Gate Thresholds
161
162| Metric | HEALTHY | DEGRADED | CRITICAL |
163|--------|---------|----------|----------|
164| DORA Score | >= 0.7 | 0.4 - 0.69 | < 0.4 |
165| SLA Compliance | >= 99% | 95 - 98.9% | < 95% |
166| Incident Severity | P3/P4/NONE | P2 | P0/P1 |
167| Defect Trend | declining/stable | stable (density > 2) | increasing + density > 5 |
168| RCA Completeness | >= 80% | 50 - 79% | < 50% |
169
170---
171
172## Report Filename Mapping
173
174| Agent | Report Filename | Step |
175|-------|----------------|------|
176| qe-metrics-optimizer | `02-dora-metrics.md` | 2 |
177| qe-defect-predictor | `03-defect-prediction.md` | 2 |
178| qe-root-cause-analyzer | `04-root-cause-analysis.md` | 2 |
179| qe-chaos-engineer | `05-chaos-resilience.md` | 4 |
180| qe-performance-tester | `06-performance-sla.md` | 4 |
181| qe-regression-analyzer | `07-regression-analysis.md` | 4 |
182| qe-pattern-learner | `08-pattern-analysis.md` | 4 |
183| Learning Persistence | `09-learning-persistence.json` | 7 |
184| qe-middleware-validator | `10-middleware-health.md` | 4 |
185| qe-sap-rfc-tester | `11-sap-health.md` | 4 |
186| qe-sod-analyzer | `12-sod-compliance.md` | 4 |
187| Feedback agents | `13-feedback-loops.md` | 8 |
188| Synthesis | `01-executive-summary.md` | 6 |
189
190---
191
192## Execution Model Options
193
194| Model | When to Use | Agent Spawn |
195|-------|-------------|-------------|
196| **Task Tool** (PRIMARY) | Claude Code sessions | `Task({ subagent_type, run_in_background: true })` |
197| **MCP Tools** | MCP server available | `fleet_init({})` / `task_submit({})` |
198| **CLI** | Terminal/scripts | `swarm init` / `agent spawn` |
199
200---
201
202## Key Principle
203
204**Production health is measured by outcomes, not intentions. This swarm provides
205evidence-based production assessment and closes the QCSD feedback loop.**
proffesor-for-testing