project-manager-agent-skill

🎯 Project Manager Agent Skill

Your AI-Powered Senior Project Manager

From unstructured brief to professional delivery — structured, evidence-backed, production-grade.

---

📋 Table of Contents

• Overview
• Problem It Solves
• Architecture
• Harness Flow — 8 Stages
• Sub-Skills
• Quality Gates
• Knowledge Pipeline
• Cross-Skill Sharing (Cluster A)
• Quick Start
• Usage Examples
• Test Suite
• Project Structure
• Configuration
• Contributing
• License

---

🌟 Overview

Project Manager Agent Skill transforms Claude into a Senior Project Manager with 15+ years of experience — certified in PMP, PRINCE2 Practitioner, and SAFe Agilist. It takes any unstructured project idea, brief, or crisis situation and produces complete, professional PM deliverables grounded in PMBOK 7th Edition, Agile Alliance, and PRINCE2 standards.

What It Produces

Deliverable	Stage	Standard
Structured Requirements Document	Stage 1	SMART Framework
Scope Statement (in/out-of-scope)	Stage 2	PMBOK 7 / PRINCE2
Stakeholder Register + Influence Grid	Stage 3	Mendelow (1991)
RACI Matrix	Stage 4	RACI Governance Standard
Risk Register (P×I Matrix)	Stage 5	ISO 31000 / PMBOK 11
Execution Timeline + Change Log	Stage 6	CPM / PRINCE2
Quality Review Report	Stage 7	PMBOK / Agile / PRINCE2
Professional Status Report (RAG)	Stage 8	Industry Standard

---

❓ Problem It Solves

Most teams lack formal PM expertise when launching initiatives. Ad-hoc project management produces:

Problem	Consequence	How This Skill Fixes It
Ambiguous scope	Feature creep	Explicit in-scope / out-of-scope boundaries + change log
Unidentified stakeholders	Late-stage opposition	Systematic stakeholder discovery + Mendelow grid
Unscored risks	Crises	Probability × Impact matrix (1–5) with mandatory mitigation for High risks
No audit trail	Scope changes go untracked	Change Management Log with escalation path
Narrative status reports	Executives can't act	Data-driven RAG (Red/Amber/Green) status reports

---

🏗 Architecture

project-manager-agent-skill/
│
├── skills/                          # Claude skill definitions
│   ├── main.md                       # Primary harness (8-stage flow)
│   ├── sub-requirements-gatherer.md  # Stage 1: Intake & SMART validation
│   ├── sub-stakeholder-mapper.md    # Stage 3: Stakeholder discovery + Mendelow grid
│   ├── sub-raci-matrix-generator.md # Stage 4: RACI with governance rules
│   ├── sub-risk-assessor.md         # Stage 5: P×I risk scoring
│   └── sub-quality-reviewer.md      # Stage 7: Multi-standard quality audit
│
├── tools/                            # Knowledge pipeline
│   ├── knowledge_updater.py         # Crawl4ai pipeline (Semantic Scholar, ArXiv, PMI)
│   └── README.md                     # Cron setup guide
│
├── tests/                            # 291 automated tests
│   ├── integration_test.py           # E2E harness validation (106 tests)
│   ├── quality_gates_test.py         # Quality gate validation (42 tests)
│   ├── partial_workflow_test.py      # Partial invocation tests (42 tests)
│   ├── test_knowledge_updater.py     # Pipeline unit tests (50 tests)
│   ├── scenario_runner.py            # 5-scenario execution (51 tests)
│   └── test-scenarios.md             # Scenario definitions
│
├── .agents/                          # Cross-skill wiring
│   ├── cross-skill-wiring.json       # Machine-readable interface spec
│   └── cross-skill-wiring.md         # Human-readable wiring docs
│
├── SECOND-KNOWLEDGE-BRAIN.md         # Self-improving knowledge base
├── CLAUDE.md                         # Skill identity & task tracking
├── PROJECT-detail.md                 # Full technical specification
├── PROJECT-DEVELOPMENT-PHASE-TRACKING.md  # Phase tracking (all 100%)
├── progression.json                  # Skill completion status
└── logs/                             # Cron output directory

---

🔄 Harness Flow — 8 Stages

The skill executes a structured 8-stage pipeline. Each stage has defined inputs, outputs, and quality gates. No stage is skipped. If a quality gate fails, the harness enters a revision loop (max 3 cycles).

┌─────────────────────────────────────────────────────────────────┐
│                    PROJECT MANAGER HARNESS                        │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  Stage 1: PROJECT INTAKE                                        │
│  │  └─► sub-requirements-gatherer                               │
│  │       Input:  Raw brief / conversation / URL                  │
│  │       Output: SMART-validated requirements document           │
│  │       Gate:   Goals are SMART; assumptions flagged            │
│  ▼                                                               │
│  Stage 2: SCOPE DEFINITION                                       │
│  │       Input:  Structured requirements                         │
│  │       Output: Scope statement (in/out, milestones, approach) │
│  │       Gate:   In-scope AND out-of-scope present               │
│  ▼                                                               │
│  Stage 3: STAKEHOLDER MAPPING                                    │
│  │  └─► sub-stakeholder-mapper                                  │
│  │       Input:  Project scope + context                        │
│  │       Output: Stakeholder register + Mendelow grid            │
│  │       Gate:   ≥5 stakeholders; all 4 quadrants addressed       │
│  ▼                                                               │
│  Stage 4: RACI MATRIX GENERATION                                 │
│  │  └─► sub-raci-matrix-generator                               │
│  │       Input:  Stakeholder register + work packages            │
│  │       Output: RACI matrix + validation report                 │
│  │       Gate:   Exactly 1 Accountable per row; no orphans       │
│  ▼                                                               │
│  Stage 5: RISK ASSESSMENT                                        │
│  │  └─► sub-risk-assessor                                       │
│  │       Input:  Scope + stakeholder map + constraints           │
│  │       Output: Risk register (P×I scored) + heat map          │
│  │       Gate:   ≥5 risks; High (≥12) risks have mitigation     │
│  ▼                                                               │
│  Stage 6: EXECUTION TRACKING                                     │
│  │       Input:  Scope + RACI + Risk register                   │
│  │       Output: Timeline + milestones + change log template     │
│  │       Gate:   ≥3 milestones; change log template present      │
│  ▼                                                               │
│  Stage 7: QUALITY REVIEW                                         │
│  │  └─► sub-quality-reviewer                                    │
│  │       Input:  All prior deliverables                          │
│  │       Output: Quality report (pass/fail + corrections)        │
│  │       Gate:   All critical items pass                        │
│  │       ◄── REVISION LOOP (if gate fails, revise & re-check)   │
│  ▼                                                               │
│  Stage 8: STATUS REPORT                                          │
│  │       Input:  All validated outputs                          │
│  │       Output: Professional PM status report (RAG)            │
│  │       Gate:   RAG indicators present; top 3 risks listed     │
│  ▼                                                               │
│  ════════════════════════════════════════════                     │
│  ✅  COMPLETE PM DELIVERABLES PACKAGE                             │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Branching Logic

The harness supports partial invocation — users can request a single deliverable without running the full 8-stage flow:

User Request	Sub-skill Invoked	Required Inputs
"I just need a RACI matrix"	sub-raci-matrix-generator	Stakeholder register + work packages
"Create a risk register"	sub-risk-assessor	Project scope + stakeholder map + constraints
"Map the stakeholders"	sub-stakeholder-mapper	Project scope + brief
"Gather requirements"	sub-requirements-gatherer	Project brief (text, URL, or conversation)
"Review quality of my deliverables"	sub-quality-reviewer	All deliverables to review

Error Handling

Scenario	Behavior
Insufficient input	Ask up to 3 clarifying questions at once; proceed with [ASSUMPTION] flags
Conflicting requirements	Surface conflict explicitly; user decides
WebSearch/WebFetch unavailable	Use SECOND-KNOWLEDGE-BRAIN.md; flag limitation
Quality gate failure	Revision loop (max 3 cycles); present best output with warning

---

🧩 Sub-Skills

sub-requirements-gatherer.md

Purpose: Elicits, structures, and validates project requirements from any input format.

Property	Detail
Inputs	Raw brief (text/document/URL/conversation)
Outputs	SMART-validated requirements doc: goals, success criteria, constraints, assumptions, out-of-scope
Quality Gate	≥1 SMART goal; all assumptions flagged; out-of-scope populated
Standards	PMBOK 7 §2, SMART Framework (Doran, 1981)

sub-stakeholder-mapper.md

Purpose: Identifies all stakeholders, plots them on the Mendelow influence/interest grid, defines engagement strategy.

Property	Detail
Inputs	Project scope + domain context
Outputs	Stakeholder register + influence/interest grid + engagement strategies
Quality Gate	≥5 stakeholders; all 4 Mendelow quadrants; engagement strategy per stakeholder
Standards	PMBOK 7 Stakeholders Domain, Mendelow (1991)
Shared With	ba-specialist, brand-recognition-auditor (Cluster A)

sub-raci-matrix-generator.md

Purpose: Generates RACI matrix with enforced governance rules and anti-pattern detection.

Property	Detail
Inputs	Stakeholder register + work packages
Outputs	RACI matrix table + validation report
Quality Gate	Exactly 1 Accountable per row; no orphan tasks; no over-consultation
Anti-Patterns	Two Accountables, PM responsible for everything, everyone Consulted, no one Informed

sub-risk-assessor.md

Purpose: Identifies, scores, and mitigates project risks using the Probability × Impact matrix.

Property	Detail
Inputs	Project scope + stakeholder map + constraints
Outputs	Risk register (7 categories) + heat map + devil's advocate check
Quality Gate	≥5 risks; all High (≥12) have mitigation + owner; heat map populated
Standards	ISO 31000, PMBOK Chapter 11

Risk Scoring Matrix:

Impact →       1-Minimal  2-Minor  3-Moderate  4-Major  5-Critical
Probability ↓
5-Almost Certain    5       10         15          20        25
4-Likely            4        8         12          16        20
3-Possible          3        6          9          12        15
2-Unlikely          2        4          6           8        10
1-Rare              1        2          3           4         5

Thresholds:  Low 1–5  │  Medium 6–11  │  HIGH 12–25 (mandatory mitigation)

sub-quality-reviewer.md

Purpose: Validates all PM deliverables against PMBOK 7, Agile, and PRINCE2 standards. Issues pass/fail verdicts.

Property	Detail
Inputs	All deliverables from prior stages
Outputs	Quality report: score, pass/fail per item, correction notes, readiness verdict
Quality Gate	90–100% = Ready; 75–89% = Proceed with corrections; <75% = Revise
Standards	PMBOK 7th Edition, PRINCE2, Agile Alliance
Shared With	ba-specialist, brand-recognition-auditor (Cluster A)

---

🛡 Quality Gates

8 mandatory quality gates enforced before final output delivery:

#	Gate	Rule	Failure Action
1	SMART Goals Gate	All goals pass SMART validation	Rewrite as SMART; flag originals
2	Scope Completeness Gate	In-scope AND out-of-scope defined	Require explicit sections
3	Stakeholder Coverage Gate	≥5 stakeholders; all 4 Mendelow quadrants	Flag empty quadrants as risk
4	RACI Integrity Gate	Exactly 1 Accountable per work package; no orphans	Resolve ownership
5	Risk Sufficiency Gate	≥5 risks; all High (≥12) have mitigation + owner	Add mitigation plans
6	Timeline Gate	≥3 milestones with dates and owners	Require milestone detail
7	Change Governance Gate	Change log template + escalation path present	Add governance infrastructure
8	Quality Review Gate	All critical items pass before Status Report	Revision loop (max 3 cycles)

---

🧠 Knowledge Pipeline

The skill includes a self-improving knowledge pipeline that crawls authoritative PM sources and appends findings to SECOND-KNOWLEDGE-BRAIN.md.

Sources

Source	Type	Frequency
Semantic Scholar API	Academic papers (PM, risk, stakeholder)	Weekly
ArXiv API	Preprints (cs.SE, cs.AI)	Weekly
PMI Learning Library	PM research & reports	Weekly
Agile Alliance Resources	Agile practices & case studies	Weekly

Features

• Deduplication: SHA-256 hash of DOI/URL ensures no duplicate entries
• Relevance scoring: Domain keyword matching (title = 2pts, abstract = 1pt)
• Atomic writes: Temp file → rename pattern prevents partial updates
• Rate limiting: 1.5s delay between API calls; 3 retries with timeout
• Dry-run mode: Preview entries before writing

Setup

# Manual run
python tools/knowledge_updater.py

# Dry run (preview only)
python tools/knowledge_updater.py --dry-run

# Weekly cron (Linux/macOS)
0 6 * * 1 cd /path/to/project-manager-agent-skill && python3 tools/knowledge_updater.py >> logs/knowledge_updater.log 2>&1

# Windows Task Scheduler: see tools/README.md for full instructions

---

🔗 Cross-Skill Sharing (Cluster A)

This skill is part of Cluster A (Professional Role & Audit Harnesses) and shares 3 sub-skills across the cluster:

┌──────────────────────────────────────────────────────┐
│                    CLUSTER A                          │
│                                                      │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐│
│  │ project-     │  │ ba-          │  │ brand-       ││
│  │ manager      │  │ specialist   │  │ recognition- ││
│  │              │  │              │  │ auditor      ││
│  │  ✦ stake-   │──│  ← stake-   │  │  ← stake-   ││
│  │    holder    │  │    holder    │  │    holder    ││
│  │    mapper    │  │    mapper    │  │    mapper    ││
│  │              │  │              │  │              ││
│  │  ✦ require- │──│  ← require- │  │              ││
│  │    ments     │  │    ments     │  │              ││
│  │    gatherer │  │    gatherer  │  │              ││
│  │              │  │              │  │              ││
│  │  ✦ quality  │──│  ← quality  │──│  ← quality  ││
│  │    reviewer │  │    reviewer  │  │    reviewer  ││
│  └──────────────┘  └──────────────┘  └──────────────┘│
│                                                      │
│  Shared: 8 quality gate standard across all skills   │
└──────────────────────────────────────────────────────┘

Shared Sub-Skill	Shared With	Adaptation
sub-stakeholder-mapper	ba-specialist, brand-recognition-auditor	BA: requirements stakeholders; Brand: brand stakeholders
sub-requirements-gatherer	ba-specialist	BA: business requirements + functional specifications
sub-quality-reviewer	ba-specialist, brand-recognition-auditor	BA: BA checklist; Brand: brand audit checklist

---

🚀 Quick Start

Prerequisites

• Python 3.10+
• requests library (pip install requests)
• Optional: crawl4ai for web scraping (pip install crawl4ai)

Installation

# Clone the repository
git clone https://github.com/dungnotnull/project-manager-agent-skill.git
cd project-manager-agent-skill

# Install dependencies
pip install requests

# Optional: for web scraping in knowledge pipeline
pip install crawl4ai

Using with Claude Code CLI

# Full harness invocation
/project-manager "We're launching a B2B SaaS tool for HR teams. Team of 6, 6-month runway."

# Partial invocation — just the RACI matrix
/project-manager "I need a RACI for our data warehouse migration. Work packages: [list]. Stakeholders: [list]."

# Partial invocation — just the risk register
/project-manager "Create a risk register for our legacy ERP migration to AWS."

# Cold start (no brief provided)
/project-manager

Running Tests

# Run all test suites (291 tests)
python tests/integration_test.py --verbose
python tests/quality_gates_test.py --verbose
python tests/partial_workflow_test.py --verbose
python tests/test_knowledge_updater.py --verbose
python tests/scenario_runner.py --verbose

# Run a single scenario
python tests/scenario_runner.py --scenario 1

---

📖 Usage Examples

Example 1: Full Project Launch (SaaS Startup)

Input:

"We're launching a B2B SaaS tool for HR teams to automate employee onboarding. We have a team of 6, a 6-month runway, and need to go live by Q4."

Output: Complete PM deliverables package — SMART goals, scope (in/out), 6+ stakeholders mapped on Mendelow grid, RACI matrix, risk register with 5+ scored risks, timeline with 4 milestones, change log template, quality review, and professional status report with RAG indicators.

Example 2: Partial — RACI Only

Input:

"I just need a RACI matrix for our data warehouse migration. Work packages: design, build, test, deploy. Stakeholders: Data Engineer, Analytics Lead, DBA, Sponsor."

Output: RACI matrix only — no scope, no risk register, no status report. Governance checks included (single-A rule, no orphans).

Example 3: Crisis Recovery

Input:

"Our project is already 3 months in and we're behind. We've added 4 new features since kickoff. The team is overwhelmed."

Output: RED status report, scope creep quantified (4 additions), 3+ HIGH risks (schedule overrun, burnout, quality degradation), recovery options presented as explicit choices (not prescribed), change log with 4 documented additions.

Example 4: Regulated Industry (Healthcare)

Input:

"We're building a patient portal for a hospital. Must integrate with Epic EHR. 9 months. Must comply with HIPAA."

Output: HIPAA compliance as hard constraint (not just a risk), compliance officer in "Manage Closely" quadrant, data breach risk scored Critical (5×5=25), penetration testing as in-scope deliverable, separate Compliance RAG status in report.

---

✅ Test Suite

291 automated tests — all passing.

Test Suite	Tests	Coverage
integration_test.py	106	E2E harness flow, all 8 stages, sub-skill references, quality gates
quality_gates_test.py	42	All 8 quality gates: SMART, scope, stakeholder, RACI, risk, timeline, change, quality
partial_workflow_test.py	42	RACI-only, risk-only, stakeholder-only, requirements-only, quality-only, error handling
test_knowledge_updater.py	50	Deduplication, relevance scoring, Semantic Scholar parser, ArXiv parser, atomic writes
scenario_runner.py	51	5 scenarios: SaaS launch, legacy migration, RACI-only, scope creep crisis, healthcare

Scenario Results

#	Scenario	Tests	Result
1	New SaaS Startup Project	17/17	✅ Pass
2	Legacy System Migration (IT Modernization)	8/8	✅ Pass
3	Partial Workflow — RACI Only	8/8	✅ Pass
4	Crisis Recovery — Pre-existing Scope Creep	8/8	✅ Pass
5	Regulated Industry — Healthcare IT Project	10/10	✅ Pass

---

📂 Project Structure

project-manager-agent-skill/
│
├── CLAUDE.md                              # Skill identity, persona, task tracking
├── PROJECT-detail.md                      # Full technical specification
├── PROJECT-DEVELOPMENT-PHASE-TRACKING.md  # Phase-by-phase build tracking (all 100%)
├── SECOND-KNOWLEDGE-BRAIN.md              # Self-improving knowledge base (auto-updated)
├── progression.json                       # Skill completion status
├── idea.txt                               # Original project concept
│
├── skills/                                # Claude skill definitions
│   ├── main.md                            # Primary 8-stage harness + branching + error handling
│   ├── sub-requirements-gatherer.md       # Stage 1: Intake, SMART validation
│   ├── sub-stakeholder-mapper.md         # Stage 3: Stakeholder discovery, Mendelow grid
│   ├── sub-raci-matrix-generator.md      # Stage 4: RACI with governance + anti-patterns
│   ├── sub-risk-assessor.md              # Stage 5: P×I risk scoring, 7 categories
│   └── sub-quality-reviewer.md           # Stage 7: Multi-standard quality audit
│
├── tools/                                 # Knowledge pipeline
│   ├── knowledge_updater.py              # Semantic Scholar + ArXiv + web scrape pipeline
│   └── README.md                          # Cron/scheduler setup guide
│
├── tests/                                 # 291 automated tests
│   ├── integration_test.py               # 106 E2E tests
│   ├── quality_gates_test.py              # 42 quality gate tests
│   ├── partial_workflow_test.py           # 42 partial invocation tests
│   ├── test_knowledge_updater.py          # 50 pipeline unit tests
│   ├── scenario_runner.py                 # 51 scenario tests
│   └── test-scenarios.md                  # Scenario definitions
│
├── .agents/                               # Cross-skill wiring (Cluster A)
│   ├── cross-skill-wiring.json            # Machine-readable interface spec
│   └── cross-skill-wiring.md              # Human-readable wiring docs
│
└── logs/                                  # Cron output directory

---

⚙️ Configuration

Knowledge Pipeline

Setting	Default	Description
MIN_RELEVANCE_SCORE	2	Minimum relevance score for inclusion
RATE_LIMIT_SECONDS	1.5	Delay between API calls
REQUEST_TIMEOUT	20s	HTTP request timeout
MAX_RETRIES	3	Retry attempts per API call
DOMAIN_KEYWORDS	24	Keywords for relevance scoring

Intake Questions (5 Core)

#	Question	Covers
1	What is the project goal and why does it matter?	Goals, business justification
2	What is the deadline or target go-live date?	Timeline, urgency
3	Who are the main stakeholders and their roles?	Stakeholder identification
4	What are the known constraints (budget, technology, team)?	Resources, limitations
5	What does success look like?	Success criteria, KPIs

Industry-specific extensions are triggered automatically when regulated domains are detected (healthcare/HIPAA, finance/SOX, government/procurement).

---

🤝 Contributing

Contributions are welcome! Here's how:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/your-feature)
3. Commit your changes (git commit -m 'Add your feature')
4. Push to the branch (git push origin feature/your-feature)
5. Open a Pull Request

Areas for Contribution

• 🌐 Additional knowledge sources (IEEE, Scopus, Springer)
• 🔧 New sub-skills (budget planner, resource allocator)
• 🧪 More test scenarios (construction, government, non-profit)
• 🌍 Internationalization (multi-language PM frameworks)
• 📊 Dashboard/visualization templates

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Built with ❤️ for project managers who demand structure, evidence, and accountability.

Standards: PMBOK 7th Edition · PRINCE2 · Agile Alliance · ISO 31000

⬆ Back to Top