Example validation check

🎯 Architectural Purpose

In next-generation enterprise environments, PMO documentation and governance checks are increasingly handled by autonomous single-agent and multi-agent systems. Instead of human practitioners manually copying data across dozens of templates, autonomous software agents: 1. Discover required project artifacts dynamically. 2. Ingest meeting notes and architectural decisions. 3. Traverse document dependency directed acyclic graphs (DAGs). 4. Synthesize, validate, and self-correct project deliverables before requesting executive human sign-off.

This guide provides technical reference blueprints for integrating Tasleemat into Single-Agent MCP Runtimes and Multi-Agent Orchestration Swarms.

flowchart TD
    USER["Project Sponsor / PM Request"] --> ORCH["<b>Orchestrator Agent (Lead PM)</b><br/>Determines Tier & Pack via get_tailoring_profile()"]

    subgraph Swarm["Specialized Multi-Agent Swarm"]
        ORCH -->|Assigns Charter & Scope| BA["<b>Business Analyst Agent</b><br/>Drafts PMO-01.01, 03.01, 04.02.05"]
        ORCH -->|Assigns WBS & Timeline| SCHED["<b>Schedule Engineer Agent</b><br/>Drafts PMO-04.02.06, 04.03.08"]
        ORCH -->|Assigns Risk & Contingency| RISK["<b>Risk Specialist Agent</b><br/>Drafts PMO-04.08.02 Risk Matrix"]
    end

    BA --> GATEKEEPER["<b>Gatekeeper QA Agent</b><br/>Runs tasleemat.validate_artifact_data(strict=True)"]
    SCHED --> GATEKEEPER
    RISK --> GATEKEEPER

    GATEKEEPER -->|Validation Errors| RETRY["Self-Correction Feedback Loop"]
    RETRY --> Swarm

    GATEKEEPER -->|100% Schema Valid| HUMAN["<b>Human-in-the-Loop Gate Approval</b><br/>Rendered HTML via tasleemat.render_artifact()"]

🔌 1. Single-Agent Integration via Model Context Protocol (MCP)

For autonomous coding assistants and single-agent environments (such as Claude Desktop, Cursor, OpenCode, and Antigravity), Tasleemat provides an official Model Context Protocol server.

Connecting via Stdio Configuration

Add Tasleemat to your agent client configuration file:

{
  "mcpServers": {
    "tasleemat": {
      "command": "tasleemat",
      "args": ["mcp"]
    }
  }
}

The 6 Core Exposed MCP Tools

Tool Name	Parameters	Machine Role
`tasleemat_list_artifacts`	`category?: string`, `tier?: string`	Discovers available deliverable codes across the 8 lifecycle phases.
`tasleemat_get_artifact`	`artifact_id: string`	Retrieves canonical metadata, title, and schema definition.
`tasleemat_get_asset`	`artifact_id: string`, `role: string`, `lang?: string`	Fetches template, prompt, guide, schema, or example files.
`tasleemat_get_lexicon`	`category?: string`, `query?: string`	Resolves official bilingual PMO terminology to prevent semantic drift.
`tasleemat_validate_data`	`artifact_id: string`, `data: object`, `mode?: "strict" \\| "draft"`	Validates structured payload against Draft 2020-12 companion schema.
`tasleemat_render_artifact`	`artifact_id: string`, `data?: object`, `lang?: string`	Renders validated deliverable to standalone styled HTML.

👥 2. Multi-Agent PMO Swarm Blueprint (CrewAI / LangGraph)

When coordinating complex delivery across multi-agent frameworks, divide responsibilities across four specialized agent personas:

A. The Orchestrator Agent (Virtual PMO Director)

Role: Analyzes user requirements, resolves the tailoring tier and methodology pack using tasleemat.get_tailoring_profile(tier, pack), and sequences task dependencies according to docs/en/07_document_dependencies.md.
System Prompt Directive:

"You are the Lead PMO Orchestrator. You manage the project DAG. Never allow downstream deliverables to be drafted before upstream prerequisites have passed schema validation."

B. The Business Analyst Agent

Role: Synthesizes PMO-01.01 (Business Case), PMO-03.01 (Project Charter), and PMO-04.02.05 (Scope Statement).
Tool Access: get_artifact_asset(role="prompt"), get_artifact_asset(role="schema").

C. The Risk & Governance Agent

Role: Formulates the Risk Register (PMO-04.08.02) and AI Model Card (PMO-02.05), ensuring strict probability/impact matrix alignment.
Tool Access: get_artifact_asset(role="dictionary"), validate_artifact_data.

D. The Gatekeeper / QA Auditor Agent

Role: Programmatic gate enforcer. Intercepts draft deliverables, parses embedded data tables against the 114 Draft 2020-12 companion schemas, and either approves or issues structured remediation directives.

🐍 3. LangGraph / Python Multi-Agent Coordination Code Sample

import tasleemat
from typing import Dict, Any

class PMOGatekeeperNode:
    """Automated Quality Gatekeeper node in an agentic pipeline."""

    def __init__(self, artifact_id: str):
        self.artifact_id = artifact_id

    def evaluate_agent_output(self, generated_payload: Dict[str, Any]) -> Dict[str, Any]:
        # Perform strict Draft 2020-12 schema validation
        report = tasleemat.validate_artifact_data(
            artifact_id=self.artifact_id,
            data=generated_payload,
            strict=True
        )

        if report["valid"]:
            # Render deliverable to HTML for human review
            html = tasleemat.render_artifact(
                artifact_id=self.artifact_id,
                data=generated_payload,
                format="html"
            )
            return {
                "status": "APPROVED",
                "html_preview": html,
                "remediation_instructions": None
            }
        else:
            # Generate structured remediation feedback for drafting agent
            return {
                "status": "REJECTED",
                "html_preview": None,
                "remediation_instructions": {
                    "errors": report["errors"],
                    "guidance": "Correct payload according to Draft 2020-12 schema violations."
                }
            }

# Example validation check
gatekeeper = PMOGatekeeperNode("PMO-03.01")
agent_payload = {
    "project_name": "Cloud Data Hub",
    "project_code": "PRJ-2026-001",
    "start_date": "2026-11-01",
    "sponsor": "Executive Sponsor"
}
result = gatekeeper.evaluate_agent_output(agent_payload)
print(f"Gatekeeper Outcome: {result['status']}")

🔁 4. The Agentic Self-Correction Loop

When an autonomous drafting agent produces invalid schema outputs, your architecture must not crash or ask a human for help. Instead, implement the Self-Correction Feedback Loop:

Step 1: Agent calls tasleemat_validate_data(artifact_id, payload, mode="draft")
Step 2: Validator returns {"valid": false, "errors": [{"path": "/start_date", "message": "Expected format: date"}]}
Step 3: Agent catches errors and re-prompts itself:
        "Field '/start_date' violated RFC 3339 format. Convert date value to YYYY-MM-DD."
Step 4: Agent updates payload and re-validates in strict mode.
Step 5: Only upon {"valid": true} does the agent write output to the workspace.