
ahmedemad3/super-ai-agency-framework
Developer tools🚀 A persona-driven AI development framework that merges with and to deliver a structured, gated, spec-driven development workflow from strategy to production with a full post-ship learning loop.
Обзор
🚀 A persona-driven AI development framework that merges with and to deliver a structured, gated, spec-driven development workflow from strategy to production with a full post-ship learning loop. The Super AI Agency Framework is a that follows a strict, gated, 9-phase workflow. Instead of vibe-coding, every feature goes through mandatory phases — from strategy to production and post-ship learning — with specialized AI personas responsible for each phase. Addressing all gaps identified in the Agency + Spec-Kit integration analysis. : 🏛️ Agency Persona | 🔧 Spec-Kit Tool | 🔁 Bridge Auto-Sync | ✋ Human Gate | 🔧🔵 spec-kit-learn 1. Open your AI coding agent in the project directory 2. Say: or 3. The Orchestrator detects no artifacts → begins Phase 0 4. Follow the prompts through each phase 5. The bridge sync runs — you do not need to trigger it manually The complete step-by-step playbook for running any project through the 9-phase framework.
README
Super AI Agency Framework
🚀 A persona-driven AI development framework that merges 13 expert AI agents with GitHub Spec-Kit v0.4.4 and spec-kit-learn v1.1.0 to deliver a structured, gated, spec-driven development workflow from strategy to production with a full post-ship learning loop.
Table of Contents
- What is This?
- 🆕 What’s New in v2.0
- The 9-Phase Workflow
- 🗺️ Workflow Diagram
- Get Started
- Simulation Guide
- Complete Step-by-Step Workflow
- AI Agent Roster
- Slash Commands
- Project Structure
- Constitution & Governance
- Spec-Kit Extensions
- Extension Integration Guide
- Bridge Script
- Emergency Recovery
- Extending the Framework
- Compatibility
- Contributing
- License
🤔 What is This?
The Super AI Agency Framework is a complete AI-powered software development team that follows a strict, gated, 9-phase workflow. Instead of vibe-coding, every feature goes through mandatory phases — from strategy to production and post-ship learning — with specialized AI personas responsible for each phase.
Why Not Just Use Spec-Kit or Copilot Alone?
| Feature | Spec-Kit Alone | Copilot/Agent Alone | Super AI Agency v2 |
|---|---|---|---|
| Structured spec-driven workflow | ✅ | ❌ | ✅ |
| Specialized AI personas | ❌ | ❌ | ✅ 13 experts |
| Gated phase transitions | ❌ | ❌ | ✅ User approval |
| 7-Lens + 5-pass code review | ❌ | ❌ | ✅ Dual review |
| Test-first development (mandatory) | Partial | ❌ | ✅ Constitutionally enforced |
| Production deployment pipeline | ❌ | ❌ | ✅ DevOps |
| Cross-artifact consistency (5 artifacts) | Partial | ❌ | ✅ Dual-track |
| Constitution enforcement | ✅ | ❌ | ✅ XML per-protocol |
| Automatic bridge sync | ❌ | ❌ | ✅ Every approval |
| Educational learning guides | ❌ | ❌ | ✅ spec-kit-learn |
| Visual architecture diagrams | ❌ | ❌ | ✅ Auto-generated |
| Post-ship retrospective loop | ❌ | ❌ | ✅ Mandatory Phase 8 |
| Git sprint branching | ❌ | ❌ | ✅ Per sprint |
| Constitution self-improvement | ❌ | ❌ | ✅ Retro-driven |
🆕 What’s New in v2.0
Addressing all gaps identified in the Agency + Spec-Kit integration analysis.
🔴 Critical Gaps Fixed
| # | Gap Fixed | How |
|---|---|---|
| G-01 | Bridge sync after STRATEGY.md | Now automatic in Orchestrator `` tag |
| G-02 | Bridge sync after BRD.md | Now automatic in Orchestrator `` tag |
| G-03 | Bridge sync after SAD.md | Now automatic in Orchestrator `` tag |
| G-04 | Bridge sync after SCHEMA.sql | Now automatic in Orchestrator `` tag |
| G-05 | Bridge sync after API.yaml | Now automatic in Orchestrator `` tag |
| G-06 | speckit-tasks never called |
Now mandatory in Phase 3 — generates tasks.md before backlog |
| G-07 | speckit-analyze not run post-tasks |
Now runs automatically after tasks.md is generated |
| G-08 | speckit-implement bypassed |
Now mandatory — all dev personas execute through it |
| G-09 | speckit-ship-run not called |
Now auto-triggered after production confirmed |
| G-10 | speckit-reconcile-run missing |
Now mandatory Phase 8 Step 17 |
| G-11 | Retro tools not called | Now mandatory Phase 8 Steps 18 + 19 |
⚠️ Partial Alignments Resolved
| # | Enhancement | Change |
|---|---|---|
| P-01 | speckit-verify-run optional → mandatory |
extensions.yml + Orchestrator SOP updated |
| P-02 | speckit-staff-review-run optional → mandatory |
extensions.yml + Orchestrator SOP updated |
| P-03 | speckit-analyze dual-track |
Now runs both Agency (5 artifacts) + Spec-Kit (3 artifacts) |
| P-04 | Checklist gate inconsistent | Now enforced before EVERY phase transition |
🆕 New Features
| Feature | Description |
|---|---|
| spec-kit-learn v1.1.0 | Auto-installed at Phase 0. /speckit.learn.diagrams after every plan. /speckit.learn.review after every implementation. /speckit.learn.clarify enhanced Q&A. |
| Phase 8 (Mandatory Post-Ship) | Reconcile → Retro → Retrospective-Analyze → Constitution update offer. Cannot be skipped. |
| Git Sprint Branching | sprint-[X]-[name] branch auto-created at Step 7 (Sprint Planning). PR auto-created at Phase 7 (Ship). |
| Automatic Bridge Sync | spec-kit-bridge.ps1 now runs AUTOMATICALLY after every approval via `` XML tag. Not manual. |
| Frontend Design Gate | Formalized in XML: choose Scratch vs Template before any frontend code. Auto-injects design tokens. |
| Post-Sprint Deploy Verify | After every staging and production deploy, you are explicitly asked to verify manually. |
| Constitution v2.0 | 7 new laws added to §5: automatic bridge, XML-only instructions, mandatory speckit-implement, mandatory verify, mandatory Phase 8, mandatory spec-kit-learn, git sprint branching. |
| SIMULATION_GUIDE.md | New complete step-by-step playbook for running any project through the framework. |
🌟 The 9-Phase Workflow
Phase 0: Init → /speckit.init → Scaffold workspace + install spec-kit-learn
Phase 1: Specify → /speckit.specify → PM strategy + BA BRD + Enhanced Clarify
Phase 2: Plan → /speckit.plan → SAD + DB + API + Diagrams + Consistency Check
Phase 3: Tasks → /speckit.tasks → Git branch + backlog + tasks.md + pre-impl analyze
Phase 4: Implement → /speckit.implement → TDD: QA tests first → FE/BE → learn.md → verify
Phase 5: Review → /speckit-commit → staff-review (5-pass) + 7-Lens audit + QA exec
Phase 6: Deploy → /speckit-deploy → Docker + CI/CD + Trivy + staging + manual verify
Phase 7: Ship → /speckit-ship → Regression + handover + production + PR auto-created
Phase 8: Post-Ship → /speckit.reconcile → Reconcile + Retro + Retrospective + Constitution update
Every phase requires your explicit approval before proceeding. No phase can be skipped.
Bridge sync runs AUTOMATICALLY after every approval — not a manual step.
🗺️ Workflow Diagram
flowchart TD
classDef agency fill:#1e3a5f,stroke:#4a90d9,color:#e8f4ff,font-weight:bold
classDef speckit fill:#1a3d2b,stroke:#4caf50,color:#e8f5e9,font-weight:bold
classDef bridge fill:#3d2a00,stroke:#ff9800,color:#fff3e0,font-weight:bold
classDef gate fill:#4a1942,stroke:#e040fb,color:#fce4ec,font-weight:bold
classDef human fill:#1a1a2e,stroke:#e91e63,color:#fce4ec,font-weight:bold
classDef learn fill:#1a2a3d,stroke:#00bcd4,color:#e0f7fa,font-weight:bold
START([🚀 Project Start]) --> PH0
subgraph PH0["⚙️ PHASE 0 — Init"]
P0_AG["🏛️ ORCHESTRATOR\n/speckit.init\nScaffold Artifacts/\nInstall spec-kit-learn"]:::agency
P0_SK["🔧 speckit-init +\nspeckit-checklist gate"]:::speckit
P0_AG --> P0_SK
end
PH0 --> PH1
subgraph PH1["📋 PHASE 1 — Specify & Clarify"]
P1a["🏛️ @product-manager\nSTRATEGY.md"]:::agency
P1b["🏛️ @business-analyst\nBRD.md (Gherkin, BPMN, NFRs)"]:::agency
P1br["🔁 BRIDGE AUTO-SYNC\nArtifacts/ → .specify/"]:::bridge
P1c["🔧 /speckit.learn.clarify\nEnhanced Q&A + context"]:::learn
P1d["🔧 speckit-checklist gate"]:::speckit
P1a --> P1b --> P1br --> P1c --> P1d
end
P1d --> G1{{"✋ GATE: Specs Approved?"}}:::gate
G1 -- REVISION --> P1b
G1 -- PASS --> PH2
subgraph PH2["🏗️ PHASE 2 — Plan & Design"]
P2a["🏛️ @principal-architect\nSAD.md + research.md"]:::agency
P2b["🏛️ @senior-dba\nSCHEMA.sql + SEED_DATA"]:::agency
P2c["🏛️ @backend-dev\nPHASE_X_API.yaml"]:::agency
P2br["🔁 BRIDGE AUTO-SYNC"]:::bridge
P2learn["🔧 /speckit.learn.diagrams\nComponent + System + Architecture"]:::learn
P2e["🔧 speckit-analyze\nDual-track: 5 artifacts"]:::speckit
P2a --> P2b --> P2c --> P2br --> P2learn --> P2e
end
P2e --> G2{{"✋ GATE: Consistency OK?"}}:::gate
G2 -- CRITICAL --> P2a
G2 -- PASS --> PH3
subgraph PH3["📅 PHASE 3 — Sprint Planning"]
P3git["🏛️ GIT BRANCH\nsprint-X-name created"]:::agency
P3a["🏛️ @agile-lead\nMASTER_PLAN.md"]:::agency
P3b["🔧 speckit-tasks\ntasks.md with TDD order"]:::speckit
P3br["🔁 BRIDGE AUTO-SYNC"]:::bridge
P3d["🔧 speckit-analyze\npre-impl consistency"]:::speckit
P3git --> P3a --> P3b --> P3br --> P3d
end
P3d --> G3{{"✋ GATE: Sprint Approved?"}}:::gate
G3 -- PASS --> PH4
subgraph PH4["💻 PHASE 4 — Implement (TDD Loop)"]
P4a["🏛️ @qa-engineer FIRST\nTEST_CASES written"]:::agency
P4b["🏛️ DESIGN GATE\nScratch OR Upload design"]:::agency
P4c["🔧 speckit-implement\n@frontend tasks.md"]:::speckit
P4d["🔧 speckit-implement\n@backend tasks.md + MermaidJS"]:::speckit
P4learn["🔧 /speckit.learn.review\nlearn.md auto-generated"]:::learn
P4e["🔧 speckit-verify-run\nMANDATORY — Constitution §2"]:::speckit
P4a --> P4b --> P4c & P4d --> P4learn --> P4e
end
P4e --> G4{{"✋ GATE: Verify CRITICAL?"}}:::gate
G4 -- CRITICAL --> P4c
G4 -- PASS --> PH5
subgraph PH5["🔍 PHASE 5 — Review & Commit"]
P5a["🔧 speckit-staff-review-run\nMANDATORY 5-pass"]:::speckit
P5b["🏛️ @tech-lead\n7-Lens Audit"]:::agency
P5c["🏛️ @qa-engineer\nTest Execution >85%"]:::agency
P5a --> P5b --> P5c
end
P5c --> G5{{"✋ GATE: APPROVED?"}}:::gate
G5 -- REQUEST_CHANGES --> P4c
G5 -- APPROVED --> COMMIT["📦 git commit + push sprint branch"]
COMMIT --> LOOP{{"More tickets?"}}
LOOP -- YES --> P4a
LOOP -- NO --> PH6
subgraph PH6["🚀 PHASE 6 — Deploy to Staging"]
P6a["🏛️ @devops + speckit-deploy\nDockerfile + CI/CD + Trivy"]:::agency
P6b["✋ YOU: Manual Verify\nRun app + check browser"]:::human
P6a --> P6b
end
P6b --> G6{{"✋ GATE: Staging OK?"}}:::gate
G6 -- FAIL --> P6a
G6 -- PASS --> PH7
subgraph PH7["🏁 PHASE 7 — Ship"]
P7a["🏛️ Full Regression ALL sprints\nBLOCKER if any fails"]:::agency
P7b["🏛️ @tech-lead\nHANDOVER + WALKTHROUGH"]:::agency
P7c["✋ YOU: Approve Production"]:::human
P7d["🏛️ @devops Blue/Green"]:::agency
P7e["🔧 speckit-ship-run\nPR → main (auto)"]:::speckit
P7f["✋ YOU: Verify Production"]:::human
P7a --> P7b --> P7c --> P7d --> P7e --> P7f
end
P7f --> PH8
subgraph PH8["🔄 PHASE 8 — Post-Ship MANDATORY"]
P8a["🔧 speckit-reconcile-run\nDrift detection + fix"]:::speckit
P8br["🔁 BRIDGE AUTO-SYNC"]:::bridge
P8b["🔧 speckit-retro-run\nMetrics dashboard"]:::speckit
P8c["🔧 speckit-retrospective-analyze\nSpec adherence score"]:::speckit
P8d["🔧 speckit-constitution\nOffer update — YOUR approval"]:::speckit
P8a --> P8br --> P8b --> P8c --> P8d
end
P8d --> MORE{{"More Sprints?"}}
MORE -- YES --> PH3
MORE -- NO --> DONE(["✅ Project Complete"])
Legend: 🏛️ Agency Persona | 🔧 Spec-Kit Tool | 🔁 Bridge Auto-Sync | ✋ Human Gate | 🔧🔵 spec-kit-learn
⚡ Get Started
Prerequisites
- Python 3.11+
- uv — Python package manager
- Git
- An AI coding agent (supported list)
Installation
Option 1: Clone and Initialize (Recommended)
# Clone the repository
git clone https://github.com/ahmedemad3/super-ai-agency-framework.git
cd super-ai-agency-framework
# Install Spec-Kit CLI (if not already installed)
uv tool install specify-cli --from git+https://github.com/github/[email protected]
# Initialize Spec-Kit in the project (already done, but verify)
specify check
Option 2: Start from Scratch
# Create a new project directory
mkdir my-project && cd my-project
# Install Spec-Kit CLI
uv tool install specify-cli --from git+https://github.com/github/[email protected]
# Initialize with Spec-Kit + Agency skills
specify init . --ai agy --ai-skills --force
# Download the Agency skills (from this repo)
# Copy the .agent/ folder from this repo into your project
Option 3: Add to Existing Project
cd your-existing-project
# Initialize Spec-Kit (merge mode)
specify init . --ai agy --ai-skills --force
# Copy the Agency skills from this repo
cp -r /path/to/super-ai-agency-framework/.agent/skills/BA-skill .agent/skills/
cp -r /path/to/super-ai-agency-framework/.agent/skills/DBA-skill .agent/skills/
cp -r /path/to/super-ai-agency-framework/.agent/skills/Orchestrator-skill .agent/skills/
# ... (copy all 13 Agency skills)
# Copy governance files
cp -r /path/to/super-ai-agency-framework/Artifacts/00_Governance ./Artifacts/00_Governance/
# Copy workflows
cp -r /path/to/super-ai-agency-framework/.agent/workflows .agent/workflows/
First Run
- Open your AI coding agent in the project directory
- Say: “Start project” or
/speckit-init - The Orchestrator detects no artifacts → begins Phase 0
- Follow the prompts through each phase
- The bridge sync runs automatically — you do not need to trigger it manually
📘 Simulation Guide
The complete step-by-step playbook for running any project through the 9-phase framework.
📄 SIMULATION_GUIDE.md — Read this before starting any project. It covers:
- Every phase with your exact review checklist
- What each step produces and how to validate it
- Pass/Fail gates and when to block
- Emergency recovery commands
- Complete artifact checklist by phase
- Quick reference for all commands
📋 Complete Step-by-Step Workflow
Treat this framework as a managed engineering organization — you are the CEO, the AI is the Agency.
Phase 0: Initialize
| Step | Command | Output | Gate |
|---|---|---|---|
| 0.1 | /speckit-init |
Folder structure + Constitution loaded | “Workspace ready. What is your project idea?” |
Phase 1: Specify (Strategy + Requirements)
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 1.1 | /speckit.specify |
Product Manager | Artifacts/01_Strategy/STRATEGY.md |
⛔ Review Vision, MVP Scope, Success Metrics |
| 1.2 | (auto-chains) | Business Analyst | Artifacts/02_Specs/BRD.md |
⛔ Review User Stories + Gherkin scenarios |
Before approving: Does every story have Given/When/Then? Are NFRs specific numbers?
Phase 1b: Clarify
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 1.3 | /speckit-clarify |
Business Analyst | ## Clarifications appended to BRD.md |
⛔ Answer ALL questions with specific values |
Always run this — even if the BRD looks complete. It catches hidden assumptions.
🔗 Bridge Sync: .\.agent\scripts\spec-kit-bridge.ps1 -FeatureName "001-my-feature"
Phase 2: Plan (Architecture + Database + API)
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 2.1 | /speckit.plan |
Solution Architect | SAD.md + research.md |
⛔ Review C4 diagrams, tech stack |
| 2.2 | (auto) | Database Architect | SCHEMA.sql + SEED_DATA.sql + data-model.md |
⛔ Review normalization, indexes |
| 2.3 | (auto) | Java/Node Developer | PHASE_[X]_API.yaml |
⛔ Review endpoint design, DTOs |
⚠️ Step 2.3 is often missed — The API design is a separate step from architecture. The developer persona designs the OpenAPI contract — no implementation code yet.
🔁 Bridge Sync: Runs automatically after each approval in Phase 2.
NEW in v2.0: After architecture approval,
/speckit.learn.diagrams --allauto-generates visual architecture docs (component, system, software architecture diagrams).
Phase 2b: Analyze (Cross-Artifact Consistency)
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 2.4 | /speckit-analyze |
Tech Lead | CROSS_ARTIFACT_CONSISTENCY.md |
⛔ BLOCKER if critical inconsistencies |
Checks: BRD ↔ API mapping, Schema ↔ DTOs, SAD feasibility, naming consistency.
Phase 3: Tasks (Sprint Planning)
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 3.0 | "Plan Sprint X" |
Orchestrator | sprint-X-name git branch created |
Auto |
| 3.1 | /speckit.tasks |
Project Manager | MASTER_PLAN.md (multi-sprint overview) |
⛔ Verify MVP = Sprint 1 |
| 3.2 | (auto) | Project Manager | SPRINT_[X]_BACKLOG.md + tasks.md via speckit-tasks |
⛔ Review deps, [P] markers |
| 3.3 | (auto) | speckit-analyze | CROSS_ARTIFACT_CONSISTENCY.md (pre-impl) |
⛔ BLOCKER if CRITICAL |
NEW in v2.0:
speckit-tasksis now invoked to generatetasks.md—speckit-implementreads it for TDD order enforcement. Git branch auto-created at sprint start.
Phase 4: Implement (TDD Loop — Per Ticket)
Repeat for EACH ticket in Sprint Backlog:
| Step | Command | Persona | Output |
|---|---|---|---|
| 4.1 | /speckit.implement [TicketID] |
QA Engineer | TEST_CASES_[TicketID].md (tests written FIRST — mandatory) |
| 4.2 | (FE tickets only) | Frontend Dev | ⛔ “Scratch OR upload design reference” |
| 4.3 | (auto via speckit-implement) | Frontend Dev | Source code — reads tasks.md, marks [X] |
| 4.4 | (auto via speckit-implement) | Backend Dev | Source code + MermaidJS flowchart + sequence diagram |
| 4.5 | (auto — after_implement hook) | spec-kit-learn | learn.md — educational guide auto-generated |
| 4.6 | (auto — mandatory) | speckit-verify-run | Verification report — CRITICAL blocks Phase 5 |
NEW in v2.0:
speckit-implementis now mandatory — agents execute through it, not free-form.speckit-verify-runis now mandatory (not optional).learn.mdis auto-generated after every ticket.
Phase 5: Review (Code Quality Gate — Per Ticket)
| Step | Command | Persona | Output | Decision |
|---|---|---|---|---|
| 5.1 | (auto — mandatory) | speckit-staff-review-run | reviews/review-[timestamp].md (5-pass) |
Blocker findings → back to Phase 4 |
| 5.2 | /speckit-commit |
Tech Lead | REVIEW_[TicketID].md (7-Lens audit) |
APPROVED → Commit / REQUEST_CHANGES → Phase 4 |
| 5.3 | (auto) | QA Engineer | Test execution > 85% goal | PASS → Continue / FAIL → Phase 4 |
Review: 5-pass staff review FIRST, then 7-Lens: Architecture · Security · Resilience · Observability · Performance · Test Quality · Maintainability
NEW in v2.0:
speckit-staff-review-runis now mandatory (was optional). Must run before Tech Lead 7-Lens.
Phase 6: Deploy (Staging)
| Step | Command | Persona | Output |
|---|---|---|---|
| 6.1 | /speckit-deploy |
DevOps Engineer | Dockerfile, docker-compose.yml, CI/CD pipeline |
| 6.2 | (auto) | DevOps Engineer | Trivy scan — ⛔ BLOCKER: 0 Critical/High CVEs |
| 6.3 | (auto) | DevOps Engineer | DEPLOYMENT_LOG.md |
| 6.4 | YOU (manual) | — | Run app + verify features work in browser |
NEW in v2.0: You are explicitly asked to verify staging manually before proceeding to Phase 7.
Phase 7: Ship (Production)
| Step | Command | Persona | Output | Gate |
|---|---|---|---|---|
| 7.1 | /speckit-ship |
QA Engineer | REGRESSION_SPRINT_[X].md |
⛔ BLOCKER: ANY test fails → back to Phase 4 |
| 7.2 | (auto) | Tech Lead | SPRINT_[X]_HANDOVER.md + WALKTHROUGH.md + QUALITY_POLISH_VERIFICATION.md + CHANGELOG |
⛔ Review handover doc |
| 7.3 | YOU | — | — | ⛔ “Approve production deployment?” |
| 7.4 | (auto) | DevOps Engineer | Production deploy (Blue/Green or Canary) | Monitor error rates 15 min |
| 7.5 | (auto) | speckit-ship-run | PR created → main + changelog auto-generated | |
| 7.6 | YOU (manual) | — | — | ⛔ Verify production features work |
NEW in v2.0:
speckit-ship-runis now mandatory (was optional) — auto-creates PR with full artifact traceability.
Phase 8: Post-Ship (NEW — MANDATORY)
| Step | Tool | Output |
|---|---|---|
| 8.1 | speckit-reconcile-run |
Drift detection + surgical spec/plan/tasks fixes |
| 8.2 | speckit-retro-run |
Metrics retrospective — accuracy scores, QA rate, improvement suggestions |
| 8.3 | speckit-retrospective-analyze |
Quantitative spec adherence scoring |
| 8.4 | speckit-constitution |
Constitution update (requires YOUR explicit approval) |
NEW in v2.0: Phase 8 is now mandatory — it’s the learning loop that improves every future sprint.
Multi-Sprint Loop
After Ship, if MASTER_PLAN.md has more sprints:
→ Go back to Phase 3: "Plan Sprint [X+1]"
→ Repeat Phase 4 → 5 → 6 → 7
→ Continue until all sprints shipped
🔄 Multi-Sprint Edge Cases
- API evolution: If a later sprint introduces new endpoints, update
PHASE_[X]_API.yamlduring that sprint’s Phase 4. - Architectural changes: If a later sprint adds a major component (e.g., Elasticsearch), re-run
/speckit.planand/speckit-analyzebefore Phase 4. - Test Suite Growth: By Sprint 10, regression tests may exceed context limits. Use “critical path regression” for later sprints.
- Bridge Sync Scheduled: Always run
spec-kit-bridge.ps1after Phase 1b, Phase 2, Phase 3, AND at the end of Phase 7 to sync test cases and deliverables.
🤖 AI Agent Roster
13 Specialized Persona Skills
| # | Agent | Role | Key Output |
|---|---|---|---|
| 1 | Product Manager | Strategy & MVP Definition | STRATEGY.md |
| 2 | Technical Proposal Architect | Client-Facing Proposals | TECHNICAL_PROPOSAL.md |
| 3 | Business Analyst | Requirements & Gherkin Specs | BRD.md |
| 4 | Solution Architect | System Design (C4, 3-Pass) | SAD.md, research.md |
| 5 | Database Architect | Schema, Seeding, Performance | SCHEMA.sql, data-model.md |
| 6 | Java Developer | Spring Boot 3, Java 21+ | API implementation |
| 7 | Node.js Developer | NestJS/Express, TypeScript | API implementation |
| 8 | Frontend Developer | React/Vue/Angular | UI components |
| 9 | QA Engineer | TDD, Playwright, Regression | TEST_CASES_[ID].md |
| 10 | Tech Lead | 7-Lens Code Review | REVIEW_[ID].md |
| 11 | Project Manager | Sprint Planning, DAG Tasks | MASTER_PLAN.md |
| 12 | DevOps Engineer | Docker, CI/CD, Trivy | Dockerfile, pipelines |
| 13 | Orchestrator (CEO) | Phase Management | XML Protocol Enforcement |
12 Spec-Kit Skills (Auto-installed + spec-kit-learn)
| Skill | Command | Purpose | Status |
|---|---|---|---|
speckit-constitution |
/speckit.constitution |
Create project governing principles | Core |
speckit-specify |
/speckit.specify |
Generate feature specification | Core |
speckit-clarify |
/speckit.clarify |
Structured ambiguity resolution | Core |
speckit-plan |
/speckit.plan |
Technical implementation plan | Core |
speckit-analyze |
/speckit.analyze |
Cross-artifact consistency check (5 artifacts) | Core |
speckit-tasks |
/speckit.tasks |
Task breakdown with TDD ordering | Mandatory |
speckit-implement |
/speckit.implement |
Execute implementation (all dev personas use this) | Mandatory |
speckit-checklist |
/speckit.checklist |
Quality validation checklist | Core |
speckit-taskstoissues |
/speckit.taskstoissues |
Convert tasks to GitHub Issues | Optional |
speckit-learn-review |
/speckit.learn.review |
Educational learn.md after implementation | Auto |
speckit-learn-diagrams |
/speckit.learn.diagrams |
Visual architecture diagrams after plan | Auto |
speckit-learn-clarify |
/speckit.learn.clarify |
Enhanced clarification with context | Auto |
🔧 Slash Commands
Core Workflow Commands
| Command | Phase | What Happens |
|---|---|---|
/speckit-init |
0 | Scaffold workspace, load Constitution |
/speckit.specify |
1 | PM → BA chain: Strategy then BRD |
/speckit-clarify |
1b | BA asks 5-10 targeted questions |
/speckit.plan |
2 | Architect → DBA → API Design |
/speckit-analyze |
2b | Tech Lead cross-artifact audit |
/speckit.tasks |
3 | PM creates sprint backlog with [P] markers |
/speckit.implement |
4 | QA → FE/BE TDD implementation loop |
/speckit-commit |
5 | Tech Lead 7-Lens review + Git commit |
/speckit-deploy |
6 | DevOps: Docker → CI/CD → Staging |
/speckit-ship |
7 | Regression → Handover → Production |
spec-kit-learn Commands (Auto-triggered)
| Command | Trigger | Purpose |
|---------|---------|---------|----------|
| /speckit.learn.diagrams | After /speckit.plan | Visual architecture diagrams (Component, System, Software) |
| /speckit.learn.review | After /speckit.implement | Educational learn.md — Key Decisions, Concepts, Glossary |
| /speckit.learn.clarify | During /speckit-clarify | Enhanced Q&A with Why + Pros/Cons + Recommended option |
Enhancement Commands
| Command | Purpose |
|---|---|
/speckit.checklist |
Generate quality validation checklist |
/speckit.taskstoissues |
Push tasks to GitHub Issues |
/speckit.reconcile.run |
Post-ship drift reconciliation (Phase 8) |
/speckit.retro.run |
Sprint retrospective with metrics (Phase 8) |
/speckit.retrospective.analyze |
Spec adherence scoring (Phase 8) |
📁 Project Structure
super-ai-agency-framework/
├── .agent/
│ ├── skills/ ← 22 Skills (13 Agency + 9 Spec-Kit)
│ │ ├── Orchestrator-skill/ ← 🎯 CEO: 8-phase decision engine
│ │ ├── Product-manager-skill/ ← 📊 Strategy & MVP
│ │ ├── BA-skill/ ← 📋 Requirements (Gherkin)
│ │ ├── Solution-architect-skill/ ← 🏗️ System Design (C4)
│ │ ├── DBA-skill/ ← 🗄️ Database Design
│ │ ├── Sr.Java-skill/ ← ☕ Java 21+ Backend
│ │ ├── Sr.Node-skill/ ← 🟢 Node.js Backend
│ │ ├── Sr.Front-skill/ ← 🎨 Frontend (React/Vue/Angular)
│ │ ├── Sr.QA-skill/ ← 🧪 Testing (TDD + Regression)
│ │ ├── TL-skill/ ← 🔍 7-Lens Code Review
│ │ ├── PM-skill/ ← 📅 Sprint Planning (DAG)
│ │ ├── Sr.Devops-skill/ ← 🚀 Docker + CI/CD
│ │ ├── tech_proposal-skill/ ← 📄 Technical Proposals
│ │ └── speckit-*/ ← 🧩 9 Spec-Kit skills
│ ├── scripts/
│ │ └── spec-kit-bridge.ps1 ← 🔗 Artifact sync script
│ └── workflows/ ← ⚡ 6 workflow definitions
│ ├── speckit-init.md
│ ├── speckit-clarify.md
│ ├── speckit-analyze.md
│ ├── speckit-commit.md
│ ├── speckit-deploy.md
│ └── speckit-ship.md
├── .specify/ ← 🧩 Spec-Kit scaffold
│ ├── memory/constitution.md
│ ├── extensions/ ← Installed extensions
│ ├── presets/agency-enterprise/ ← Custom Agency preset
│ ├── scripts/
│ └── templates/
├── Artifacts/ ← 📂 Source of Truth
│ ├── 00_Governance/ ← Constitution & Artifact Map
│ ├── 01_Strategy/ ← PM output
│ ├── 02_Specs/ ← BA output
│ ├── 03_Architecture/ ← Architect output
│ ├── 04_Database/ ← DBA output
│ ├── 05_Planning/ ← Sprint backlogs
│ ├── 06_Quality_Reports/ ← Tests, reviews, regression
│ ├── 07_API_Specs/ ← OpenAPI contracts
│ └── 08_Releases/ ← Handover docs, deploy logs
├── README.md
├── LICENSE
├── CONTRIBUTING.md
├── CHANGELOG.md
└── .gitignore
🛡️ Constitution & Governance
Every AI agent is bound by the Constitution (Artifacts/00_Governance/CONSTITUTION.md). It defines:
| Section | What It Governs |
|---|---|
| 1. Quality Standards | >85% test coverage, no any types, cyclomatic complexity ≤10 |
| 2. Testing Policy | Shift-Left TDD, Happy/Negative/Security paths per ticket |
| 3. Security Baselines | OWASP Top 10, non-root Docker, no PII in logs |
| 4. Tech Stack Rules | Java 21+/Node 20+, PostgreSQL 16+, React 19/Next.js 15 |
| 5. Phase Gate Rules | User approval required, no phase skipping |
| 6. Architecture Standards | 3-Pass Refinement, C4 diagrams, API-First, DTOs only |
| 7. DevOps Standards | Multi-stage Docker, Trivy scan, Blue/Green deploy |
| 8. Documentation Standards | MermaidJS diagrams, ADRs, README per service |
The Orchestrator injects a `` into every XML protocol to enforce compliance.
🧩 Spec-Kit Extensions
7 extensions — all mandatory (no longer optional):
| # | Extension | Purpose | Command | Phase | Status |
|---|---|---|---|---|---|
| 1 | learn | Visual diagrams + educational guides + enhanced clarify | /speckit.learn.* |
1,2,4 | Mandatory + Auto |
| 2 | staff-review | 5-pass staff-engineer code review against spec | /speckit.staff-review.run |
5 | Mandatory |
| 3 | verify | Post-implementation quality gate | /speckit.verify.run |
4 | Mandatory |
| 4 | ship | Release pipeline: pre-flight, changelog, PR creation | /speckit.ship.run |
7 | Mandatory |
| 5 | reconcile | Detect & surgically fix spec drift | /speckit.reconcile.run |
8 | Mandatory |
| 6 | retrospective | Quantitative spec adherence scoring | /speckit.retrospective.analyze |
8 | Mandatory |
| 7 | retro | Sprint retro with metrics dashboard | /speckit.retro.run |
8 | Mandatory |
v2.0 Change: All extensions changed from
optional: true→optional: falseinextensions.yml. They are now enforced by the Orchestrator XML protocol.
Install additional extensions:
# Browse the catalog
specify extension search
# Install from GitHub
specify extension add --from
# List installed
specify extension list
📡 Extension Integration Guide
v2.0 Change: Extensions are now auto-triggered by the Orchestrator at specific phase gates — not manual.
| Extension | Trigger | Phase | Mandatory |
|---|---|---|---|
/speckit.learn.diagrams |
Auto after /speckit.plan approved |
2 | ✅ Yes |
/speckit.learn.clarify |
Auto during /speckit-clarify |
1b | ✅ Yes |
/speckit.learn.review |
Auto after every /speckit.implement |
4 | ✅ Yes |
/speckit.verify.run |
Auto after implementation, before Phase 5 | 4 end | ✅ Yes |
/speckit.staff-review.run |
Auto first step of /speckit-commit |
5 | ✅ Yes |
/speckit.ship.run |
Auto after production deployment confirmed | 7 end | ✅ Yes |
/speckit.reconcile.run |
Auto start of Phase 8 | 8 | ✅ Yes |
/speckit.retro.run |
Auto Phase 8 Step 2 | 8 | ✅ Yes |
/speckit.retrospective.analyze |
Auto Phase 8 Step 3 | 8 | ✅ Yes |
/speckit.constitution |
Offered by Orchestrator — YOUR approval required | 8 end | Approval only |
🔗 Bridge Script
The spec-kit-bridge.ps1 script syncs Artifacts/ (Agency source of truth) → .specify/specs/ (Spec-Kit CLI compatibility).
v2.0 Change: The bridge now runs AUTOMATICALLY after every approval via the `` XML tag in the Orchestrator SOP. You do NOT need to run it manually.
Automatic sync points (all now automatic):
- After STRATEGY.md approved →
spec.md(overwrite) - After BRD.md approved →
spec.md(append) - After Clarifications recorded →
spec.md(append) - After SAD.md approved →
plan.md - After SCHEMA.sql approved →
schema.sql+data-model.md - After API.yaml approved →
contracts/ - After SPRINT_BACKLOG approved →
tasks.md - After Phase 8 reconcile →
Artifacts/(reverse sync)
Manual use (if needed for debugging):
# Dry run (preview what would be synced)
.agent\scripts\spec-kit-bridge.ps1 -FeatureName "001-my-feature" -DryRun
# Live sync
.agent\scripts\spec-kit-bridge.ps1 -FeatureName "001-my-feature"
If bridge_sync_status shows STALE in the status block, say: “Run bridge sync for feature [name]”.
🚨 Emergency Recovery
When the AI drifts off-workflow, use these rescue commands:
| Problem | Fix |
|---|---|
| AI jumps phases | "STOP. Check ARTIFACT_MAP.md. We are missing [artifact]. Run [command]." |
| Workflow loops | Run /speckit-init to reset state detection |
| Persona identity crisis | "You are [Persona]. Load Phase [X] protocol from your SKILL.md." |
| Context window overflow | Start new conversation → paste last `` block → run /speckit-init |
| Artifact exists but low quality | Run /speckit.checklist to validate, then ask persona to regenerate |
| Need full state report | "@team-orchestrator, full state check. Read Artifacts/, compare to ARTIFACT_MAP.md. Report only, no action." |
| AI says LGTM without details | "PROTOCOL VIOLATION. Constitution requires 7-Lens tabular report. Re-run the review." |
| Tests written after code | "PROTOCOL VIOLATION. QA writes tests first. Run Step 4.1 for [TicketID]." |
🔌 Extending the Framework
The framework is designed for easy growth. Two extension paths are available:
Option 1: Add a New Agency Skill (AI Persona)
Best for: Adding a new expert role (e.g., Security Auditor, Data Engineer, Mobile Developer) Effort: ~15 minutes — just 1 file
Create a new folder in .agent/skills/ with a SKILL.md:
.agent/skills/My-New-skill/SKILL.md
SKILL.md Template:
---
name: persona-my-new-role
description: Act as a [Role] (X+ years exp) specializing in [Specialty].
---
# Role: [Title]
## Context & Experience
You have **X years of experience**. You specialize in...
## Technical Skills & Competencies
- Skill 1
- Skill 2
## 🧠 Context Loading Protocol
1. **Load Artifacts**: Read `[relevant artifacts]`
## Workflow Interaction Protocol
1. **Incoming**: You receive [trigger]
2. **Action**: [What you do]
3. **Outgoing**: Save to `Artifacts/[folder]/[FILE].md`
- End with: *"@team-orchestrator, [deliverable] is ready."*
## Deliverables
- **`FILE_NAME.md`**: Description
Then add one XML protocol block to the Orchestrator (Orchestrator-skill/SKILL.md) to wire it into a phase — done.
Option 2: Add a Spec-Kit Extension
Best for: Adding workflow commands, integrations (Jira, Linear), or quality gates Effort: 2 minutes (install) or ~30 minutes (build your own)
Install from Community
# Search available extensions
specify extension search
# Install from GitHub
specify extension add --from
This auto-registers a new skill in .agent/skills/ and adds a slash command. Zero code needed.
Build Your Own Extension
Create this structure:
.specify/extensions/my-extension/
├── extension.yml ← Extension metadata
├── commands/
│ └── speckit.my-ext.run.md ← The command prompt (AI instructions)
└── templates/ ← Optional template overrides
extension.yml:
name: my-extension
version: 1.0.0
description: What this extension does
author: Your Name
commands:
- name: speckit.my-ext.run
description: What the command does
type: agent-skill
commands/speckit.my-ext.run.md:
You are a [role]. When invoked:
1. Read [artifacts]
2. Perform [task]
3. Output results to [path]
Then install it:
specify extension add my-extension --from ./path/to/extension.zip
Quick Comparison
| New Agency Skill | Spec-Kit Extension | |
|---|---|---|
| When to use | New AI persona / expert role | New workflow command / integration |
| Files to create | 1 (SKILL.md) |
2-3 (extension.yml + command) |
| Wired into Orchestrator | Yes (add XML block) | Optional |
| Gets a slash command | No (triggered by Orchestrator) | Yes (/speckit.name.run) |
| Publishable to community | No (Agency-specific) | Yes (Spec-Kit catalog) |
| Example use cases | Security Auditor, Data Engineer, ML Engineer | Jira sync, Slack notifications, custom linters |
🤖 Compatibility
This framework works with any AI coding agent that supports Spec-Kit:
| Agent | Status | Notes |
|---|---|---|
| Antigravity (agy) | ✅ Primary | Full skills support |
| GitHub Copilot | ✅ Supported | Via /speckit.* commands |
| Claude Code | ✅ Supported | Via /speckit.* commands |
| Gemini CLI | ✅ Supported | Via /speckit.* commands |
| Cursor | ✅ Supported | Via slash commands |
| Codex CLI | ✅ Supported | Via $speckit-* skills |
| Windsurf | ✅ Supported | Via slash commands |
Contributing
See CONTRIBUTING.md for guidelines.
Changelog
See CHANGELOG.md for version history.
v2.0 (2026-04-10): Automatic bridge sync, mandatory spec-kit-learn, mandatory Phase 8 post-ship, mandatory speckit-implement and speckit-verify-run, git sprint branching, Constitution v2.0, 9-phase workflow, SIMULATION_GUIDE.md.
License
This project is licensed under the MIT License — see LICENSE for details.
Acknowledgements
- GitHub Spec-Kit — The spec-driven development toolkit this framework extends
- John Lam — Research behind Spec-Driven Development
- imviancagrace/spec-kit-learn — Educational guides extension integrated in v2.0
Рекомендуемые инструменты
Попробуйте другой запрос или уберите фильтр.
Установка
npx skillfish add ahmedemad3/super-ai-agency-framework