Verify onboarding email draft from copy agent
Check the AI-generated onboarding email against the approved tone and content brief.
AI Agents
Agent QA Verification Board
Agent QA Verification Board: Catch Bad AI Output Before It Ships
Verifying that an AI agent's output actually matches product requirements and quality standards is hard to do consistently without a system. This Agent QA Verification Board gives your team and your AI agents a shared kanban board to track every generated output from submission through spec check, quality bar review, and final approval or rejection. Built for use with the MCP server and REST API, it lets a QA agent move tasks, log findings, and keep a clear audit trail—so nothing gets approved that hasn't actually been checked.
kanboard.io/templates/agent-qa-verification
Template preview
Agent QA Verification Board
5 columns
Submitted for Review
3Holds newly generated agent outputs waiting to be checked against requirements.
Review support macro reply for refund policy
Confirm the agent's refund macro matches current policy wording exactly.
Check code comment agent output on auth module
Validate that generated code comments accurately describe the authentication logic.
Spec Check
2Verify the output satisfies the literal functional and content requirements of the task.
Confirm API doc agent covered all endpoints
Cross-reference generated API docs against the endpoint list in the REST API spec.
Validate summarizer agent hit required word count
Check that the summary agent's output falls within the 100-150 word requirement.
Quality Bar Check
3Assess tone, clarity, correctness, and polish beyond raw spec compliance.
Assess clarity of generated release notes
Read release notes agent output for jargon, ambiguity, and readability issues.
Review chatbot escalation response quality
Evaluate whether the chatbot's escalation message sounds empathetic and professional.
Check keyboard shortcuts help text accuracy
Verify the agent-generated shortcuts guide matches actual product keyboard shortcuts.
Approved
2Outputs that passed both spec and quality checks and are cleared to ship.
Ship verified command palette tutorial copy
Approved command palette walkthrough text ready for publishing to docs site.
Release confirmed macOS app changelog entry
Changelog entry for the macOS app passed both spec and quality review.
Rejected
2Outputs that failed verification and need agent rework with documented reasons.
Rework MCP server setup guide - missing steps
Agent output skipped two required configuration steps; needs regeneration.
Redo board column description - inaccurate wording
Generated column description didn't match actual kanban board behavior.
How to run this board
Step 1
New agent output lands in Submitted for Review with the task requirements attached.
Step 2
QA agent moves the task to Spec Check and verifies literal requirements are met using get_task and update_task for notes.
Step 3
Passing tasks move to Quality Bar Check for tone, correctness, and polish review.
Step 4
Tasks that pass both checks move to Approved; failing tasks move to Rejected with a documented reason.
Step 5
Rejected tasks are reworked and resubmitted to Submitted for Review to restart verification.
AI agent usage
Run it with AI agents
This playbook helps an AI QA agent systematically verify that another agent's output matches product requirements before it ships. The agent pulls tasks from the board via MCP, checks each output against acceptance criteria, and moves tasks through review states, logging discrepancies as comments so humans can audit the process.
## QA Verification Board Agent Instructions
This board has columns: `Submitted for Review`, `Spec Check`, `Quality Bar Check`, `Approved`, `Rejected`.
1. Use `list_tasks` on `Submitted for Review` to find new agent outputs awaiting verification.
2. For each task, use `get_task` to read the requirements and attached output.
3. Move the task to `Spec Check` using `move_task` while verifying the output matches stated requirements.
4. If it passes, move to `Quality Bar Check` to verify tone, correctness, and completeness against the quality bar.
5. Move to `Approved` if both checks pass, or `Rejected` with a comment explaining the gap via `update_task`.
6. Never delete tasks unless explicitly instructed; use `delete_task` only for duplicate or test entries.Pull all tasks in Submitted for Review and check them against the linked requirements doc.
Move task 'Chatbot response tone check' from Spec Check to Quality Bar Check and note any issues.
List all Rejected tasks from this week and summarize the most common failure reasons.
Learn more about the board-centric workflow in Kanban for AI Agents or open the MCP guide.
Frequently asked questions
How many QA verification boards can I run on the free plan?
The free plan includes 1 project with up to 25 tasks total, which is enough to run this 5-column QA board with its full set of sample tasks.
Can multiple AI agents work on the same board simultaneously?
Yes, since the board is accessed via the MCP server and REST API, multiple agents can read and update tasks concurrently using tools like list_tasks and update_task.
What happens if I exceed 25 tasks on the free plan?
You'll need to upgrade to the pro plan at $12/mo or $96/yr to add more tasks beyond the free plan's 25-task limit.
Does this template support keyboard shortcuts for faster review?
Yes, the kanban board supports keyboard shortcuts and a command palette for quickly moving tasks between columns during review sessions.
Can I customize the columns for a different QA workflow?
Yes, columns are fully editable. You can rename or add columns as long as you keep the total between 3 and 6 columns.
Is there a native app for managing this board on Mac?
Yes, there's a macOS app available in addition to the web interface and REST API access.
Related templates
Ready to create this board?
Sign in and use the template to create a project with columns and sample tasks.