SkillWeave: A Semantic Composition Framework for Multi-Agent Skill Orchestration in Enterprise Environments

Paper: SkillWeave: A Semantic Composition Framework for Multi-Agent Skill Orchestration in Enterprise Environments Author: Vivek Acharya (Boston University) — vacharya@bu.edu — ORCID: 0009-0002-0860-9462 Target Venue: IEEE Transactions on Services Computing

---

Overview

SkillWeave addresses a critical architectural gap in the multi-agent AI ecosystem: the absence of governed, auditable mechanisms for composing agent skills across organizational boundaries. While MCP (Model Context Protocol) standardizes agent-to-tool connectivity and A2A (Agent-to-Agent protocol) enables inter-agent communication, no existing framework handles cross-agent skill composition — ensuring that skills from different agents work together without conflicts, data leaks, or policy violations.

Core Contributions

1. Semantic Skill Composition Algebra (SSCA) — Formal model for skill compatibility, conflict detection, and dependency resolution across multi-agent boundaries
2. SkillWeave Orchestration Protocol (SWOP) — Extends MCP+A2A with skill negotiation, composition planning, and governed execution contracts
3. Organizational Skill Governance (OSG) — Integrates principal hierarchy enforcement, data boundary compliance, and regulatory policies into the skill composition lifecycle
4. Enterprise Skill Registry (ESR) — Governed marketplace with skill provenance, semantic versioning, access controls, and audit trails

---

Repository Structure

skillweave/
├── src/skillweave/          # Core framework
│   ├── __init__.py
│   ├── models.py            # Data models (Skill, Agent, Policy, Principal)
│   ├── algebra.py           # SSCA implementation (Section V)
│   ├── governance.py        # OSG implementation (Section VII)
│   ├── registry.py          # ESR implementation (Section VIII)
│   ├── protocol.py          # SWOP implementation (Section VI)
│   └── catalog.py           # 47 skills, 12 agents, 2 domains
├── experiments/
│   ├── run_experiments.py   # Main experimental harness (Section IX)
│   └── generate_tables.py   # Paper table generation (Tables I-IV)
├── results/
│   ├── raw/                 # Per-test raw CSV data
│   └── aggregated/          # Paper-ready JSON results
├── Dockerfile               # Reproducible container
├── run_experiments.sh        # One-command execution
├── requirements.txt
├── pyproject.toml
└── README.md

---

Quick Start

Prerequisites

• Python 3.9 or higher
• No external dependencies required (pure Python)

Run All Experiments

# Clone the repository
git clone https://github.com/curiosityexplorer/skillweave.git
cd skillweave

# Run experiments (seed=42 for reproducibility)
python experiments/run_experiments.py --seed 42

# Generate paper tables
python experiments/generate_tables.py

# Or LaTeX format
python experiments/generate_tables.py --format latex

Run Single Hypothesis

python experiments/run_experiments.py --hypothesis H1  # Conflict detection
python experiments/run_experiments.py --hypothesis H2  # Composition overhead
python experiments/run_experiments.py --hypothesis H3  # Policy violations
python experiments/run_experiments.py --hypothesis H4  # End-to-end reliability

Docker

docker build -t skillweave .
docker run -v $(pwd)/results:/app/results skillweave

---

Experimental Design

Testbed Configuration

Parameter	Value
Total Agents	12 (6 Financial Services, 6 Healthcare)
Total Skills	47 (24 Financial, 23 Healthcare)
Pairwise Compositions	1,081
Expert Scenarios	50
Random Seed	42
Statistical Test	Welch's t-test (α = 0.05)

Four Hypotheses

• H1: SSCA accurately detects composition conflicts → Table I
• H2: SWOP achieves acceptable negotiation overhead → Table II
• H3: OSG prevents cross-boundary policy violations → Table III
• H4: Integrated framework enables reliable composition → Table IV

Baseline Configurations

• B1 (Ungoverned): Type-level compatibility only, no policy/semantic analysis
• B2 (Static Policy): Pre-defined allow/deny rules, no dynamic negotiation
• B3 (Agent-Local): Per-agent governance, no cross-agent coordination
• SkillWeave (Full): Complete SSCA + SWOP + OSG + ESR

---

Skill Catalog

Financial Services Agents (24 skills)

Agent	Role	Skills
FS-Agent-1	Portfolio Analysis	Valuation, Allocation, Attribution, Sector Exposure
FS-Agent-2	Risk Assessment	VaR, Stress Testing, Counterparty Risk, Liquidity Risk
FS-Agent-3	Regulatory Compliance	SOX Check, AML Screening, Report Generator, Surveillance
FS-Agent-4	Trade Execution	Order Placement, Algo Trading, Pre-Trade Check, Settlement
FS-Agent-5	Audit Reporting	Trail Compilation, Control Assessment, Exceptions, Dashboard
FS-Agent-6	Client Advisory	Risk Profiling, Recommendations, Report Gen, Fee Calc

Healthcare Agents (23 skills)

Agent	Role	Skills
HC-Agent-1	Clinical Decision Support	Risk Stratification, Guideline Matching, Drug Interaction, Alerts
HC-Agent-2	Diagnostic Analysis	Lab Interpretation, Imaging, Differential Dx, Pathology
HC-Agent-3	Treatment Planning	Protocol Selection, Dosage Calc, Care Plan, Referral
HC-Agent-4	Insurance Verification	Coverage, Pre-Auth, Billing Codes, Claims
HC-Agent-5	HIPAA Compliance	PHI Audit, De-identification, Breach Detection, Reports
HC-Agent-6	Medical Records	Retrieval, Summarization, Update

---

Related Work

This paper extends the governance architecture established in:

• Governance Control Plane: github.com/curiosityexplorer/governance-control-plane
• EnterpriseAgent: github.com/curiosityexplorer/enterprise-agent

---

Citation

@article{acharya2026skillweave,
  title={SkillWeave: A Semantic Composition Framework for Multi-Agent
         Skill Orchestration in Enterprise Environments},
  author={Acharya, Vivek},
  journal={Submitted to IEEE Transactions on Services Computing},
  year={2026}
}

---