Use Case: Custom PM Agent for a Specific Project
Build a tailored project management agent based on WebPal's built-in pm_assistant, configured with project-specific reporting formats, higher security requirements, and a focused knowledge base.
Scenario
The infrastructure team is running a large cloud migration project with strict security requirements and a specific weekly reporting cadence. The built-in pm_assistant is a good starting point, but the team needs an agent configured specifically for this project: it should only have access to the project's own documents, produce status reports in a fixed format approved by the client, and never expose sensitive cost or vendor data outside the project team.
Prerequisites
- WebPal 4.2 or later
- The
pm_assistantbuilt-in agent is active on your instance - You have
agent-creatorsgroup membership (required to create agents) - The project already has a dedicated shared drive or folder
Setup
1. Review the base pm_assistant
Before creating your custom agent, spend a few minutes understanding what the built-in pm_assistant provides:
- Navigate to Flow → Agents and open pm_assistant.
- Review its system prompt, tools, and knowledge base scope.
- Note which behaviours you want to preserve and which you need to change.
Typical things you'll want to customise: the reporting template, the knowledge base scope, who can access the agent, and whether it can write to project documents.
2. Create the custom agent
- In Flow → Agents, click New Agent.
- Name it clearly — for example, CloudMigration_PM.
- In the Based on field, select pm_assistant. This imports the base configuration as a starting point.
- Set Visibility to Team and select only the project team members. This prevents anyone outside the team from accessing the agent or its knowledge.
3. Restrict the knowledge base
- In the agent's Knowledge Base settings, remove the default company-wide scope.
- Add only the Cloud Migration Project folder (e.g. Projects / Cloud Migration 2026).
- Explicitly exclude the Costs & Contracts sub-folder if it contains vendor pricing that should not be surfaced in general responses.
4. Customise the system prompt for project-specific reporting
Replace the base system prompt with one tailored to the project's reporting format:
1. Answering questions about project status, milestones, risks, and action items based on the project documents in your knowledge base.
2. Generating weekly status reports using the following format:
• Executive Summary (2–3 sentences)
• Milestones: Completed This Week / Upcoming Next Week
• Risks & Issues (RAG status: Red / Amber / Green)
• Action Items (owner, due date)
• Budget note: flag only if spend is outside agreed tolerance
3. Never include vendor contract values, individual salary data, or information from the Costs & Contracts folder in responses.
Always be concise and factual. If the answer is not in your knowledge base, say so clearly.
5. Configure tools
Enable the following tools, and disable everything else:
| Tool | Purpose | Enabled |
|---|---|---|
document_search |
Search the project knowledge base | Yes |
document_read |
Read document content for context | Yes |
document_write |
Save status reports to the project folder | Yes (restricted to project folder) |
user_notify |
Send notifications to team members | Optional |
external_api |
Query external systems (e.g. Jira, Azure DevOps) | Disable unless specifically needed |
6. Test the agent
- Open a chat with CloudMigration_PM.
- Ask it to summarise the current project status.
- Ask it to generate a weekly status report.
- Verify it does not surface restricted content (test with a question about vendor pricing).
- If anything is wrong, adjust the system prompt or knowledge base scope and re-test.
Daily Use
Checking project status
The agent searches the risk register and current status documents, then returns a summarised list with RAG ratings.
Generating the weekly status report
The agent drafts the report in the approved format and saves it as a new document in the project folder.
Tracking action items
Result
The project team has a dedicated AI assistant that knows the project inside out — without any risk of leaking sensitive data to other teams. Reports are consistently formatted, generated in seconds, and automatically filed. The team can build on the same pattern for any future project by spinning up a new custom agent from pm_assistant in minutes.