Summary

Prompt locking pins a library prompt to an agent so that workflow designers cannot change it. This guide walks through the full process: creating the prompt in the Prompt Library, assigning it to an agent, and toggling the Lock Prompt switch.

The steps

1
Open the Prompt Library
Navigate to Agents in the sidebar. The Agents page has two tabs at the top: Agents and Prompt Library. Click the Prompt Library tab. This is where all reusable prompt templates for your tenant are stored and managed.
If no prompts exist yet, you will see an empty state. You will create a new prompt in the next step. If the prompt you want to lock already exists in the library, skip to Step 3.
Do this
- Go to Agents in the sidebar.
- Click the Prompt Library tab.
- Check whether the prompt you need already exists. If it does, skip to Step 3.
Example
A content team has a brand-voice prompt they use in three different workflows. They add it to the Prompt Library once, then lock it to each agent that needs it, rather than duplicating the prompt text in each agent's inline configuration.
Best practice
- Use the Prompt Library for any prompt that will be reused across more than one agent or workflow. Inline prompts are fine for one-off configurations, but library prompts are the right foundation for locking.
- You need the Admin or Create role to create and edit prompts. The Curate role can view the Prompt Library but cannot create or edit prompts.
2
Create the prompt in the Prompt Library
Click New Prompt to open the creation form. Fill in the required and optional fields:
- Name (required): a descriptive label for the prompt, such as "Brand Voice Reviewer" or "Compliance Checker v2". This name appears in the agent edit form's library prompt selector, so clarity here matters.
- Description (optional): a brief explanation of what the prompt does and when to use it.
- Category: select from Content Generation, Review, Safety, Analysis, Transformation, or Custom.
- Tags (optional): comma-separated labels for filtering and organization.
- Content: the prompt text itself. Write it in the Content text area, or click AI Assistant to generate a draft. If you use AI Assistant, enter a narrative description of what you want the prompt to accomplish, click Generate, review the result, and edit as needed before saving.
- Agent (optional): you can attribute the prompt to a specific agent here, but this is not required for locking. You will assign and lock the prompt from the agent edit form in Step 4.
Click Create to save the prompt. It now appears in the Prompt Library and is available for assignment to any agent.
Do this
- Click New Prompt.
- Enter a clear, descriptive Name.
- Select the appropriate Category.
- Write or generate the prompt Content.
- Click Create.
Example
A safety team creates a prompt named "Output Safety Screener" in the Safety category with the tag safety. The prompt instructs the agent to check output for prohibited content categories before passing it downstream. Once saved, they can lock this prompt to every safety-review agent in the tenant.
Best practice
- Write the prompt content as if it will be read by a fresh agent with no other context: be explicit about the agent's role, the task, and the output format you expect.
- Use the AI Assistant for a first draft, then edit for precision. Generated prompts are a starting point, not a finished product.
- Name prompts by function and version if you expect to iterate: "Brand Voice Reviewer v1" is easier to manage than "Brand Voice Reviewer" when v2 exists alongside it.
3
Open the agent you want to lock the prompt to
Click the Agents tab to return to the agent list. Find the agent you want to configure and click its name to open the agent detail page. On the detail page, click the edit action to open the agent edit form.
If the agent does not exist yet, create it first: click New Agent on the agent list page, fill in the Name (required), optional Description and Tags, and click Create Agent. Then open the new agent's detail page and proceed to edit.
Do this
- Click the Agents tab.
- Click the agent's name in the list to open its detail page.
- Open the agent edit form.
- If the agent does not exist, click New Agent, fill in the form, click Create Agent, then open the new agent to edit.
Example
The team has an existing agent named "Compliance Checker". They click its name in the agent list, open the detail page, and click the edit action to reach the edit form where the prompt source options are available.
Best practice
- Only Admin and Create roles can edit agents. If you do not see an edit option, ask your administrator to assign you the appropriate role under Admin → Settings → Users and Roles.
- Check the agent's current versions before editing. If the agent is already in use in production workflows, consider whether a prompt change should be accompanied by a new agent version rather than an edit to the existing configuration.
4
Select a Library Prompt and toggle Lock Prompt
On the agent edit form, locate the prompt source selector. You will see two options:
- Inline Prompt: write a prompt directly on the agent.
- Library Prompt: select a saved prompt from the Prompt Library.
Select Library Prompt. A dropdown or selector appears listing the prompts available in your Prompt Library. Select the prompt you created in Step 2 (or the existing prompt you want to lock).
After selecting the library prompt, the Lock Prompt toggle appears on the form. Toggle it on.
Save the agent. The lock is now active.
Do this
- On the agent edit form, select Library Prompt as the prompt source.
- Choose the target prompt from the library selector.
- Toggle Lock Prompt on.
- Save the agent.
Example
After selecting "Output Safety Screener" from the library, the Lock Prompt toggle appears. The team switches it on and saves. From this point, any workflow that uses this agent will always run with the "Output Safety Screener" prompt. The prompt selector in the Workflow Editor is greyed out for this agent.
Best practice
- Lock Prompt only appears after you select a Library Prompt. If you do not see it, confirm you have switched from Inline Prompt to Library Prompt mode.
- Communicate to your workflow designers that the prompt is locked and why. The greyed-out selector in the Workflow Editor is self-explanatory, but a note in the agent's Description field explaining the lock rationale helps future editors understand the intent.
5
Verify the lock in the Workflow Editor
To confirm the lock is working, open any workflow that uses this agent (or create a test workflow and add the agent to an Agent Interaction node). In the Workflow Editor, select the Agent Interaction node that references the locked agent and open the Properties Panel.
The prompt selector for this agent should be greyed out. Workflow designers can see which prompt is assigned but cannot change it. This confirms the lock is active and the agent will use the locked prompt in every workflow instance.
Do this
- Open a workflow that uses the locked agent in the Workflow Editor.
- Select the Agent Interaction node.
- Open the Properties Panel and locate the prompt selector.
- Confirm the prompt selector is greyed out and shows the locked prompt name.
Example
A workflow designer opens the "Content Review" workflow in the Workflow Editor and clicks the Agent Interaction node that uses the "Compliance Checker" agent. The Properties Panel shows "Output Safety Screener" in the prompt field, greyed out. They cannot change it. The lock is confirmed.
Best practice
- Test the lock verification step with a user who has the Curate role, since Curate users can view the Workflow Editor and agent configurations but cannot modify agents. Their view is the closest to what a workflow designer without agent-edit permissions will see.
- If the prompt selector is not greyed out, return to the agent edit form and confirm the Lock Prompt toggle is on and the agent was saved after toggling.

What prompt locking is and why it matters

Prompt locking is a governance control. It answers a specific problem: when an AI agent's behavior needs to be consistent across every workflow that uses it, how do you prevent a workflow designer from accidentally or deliberately overriding the prompt?

In DAVE, agents are reusable components. A single agent can be referenced by multiple workflows. Without prompt locking, each workflow that uses the agent could theoretically use a different prompt, producing different behavior from what appears to be the same agent. For compliance-sensitive agents (safety screeners, output validators, tone checkers), that variability is a governance risk.

When a library prompt is locked to an agent, the prompt selector for that agent is greyed out in the Workflow Editor. The lock is enforced at the platform level: it is not a convention or a documentation note. Workflow designers cannot override it regardless of their role. The agent always uses its locked prompt.

The result is a single place to define the instruction and a single switch to protect it. When the prompt needs to change, the change happens in the Prompt Library. When the prompt needs to be frozen, the lock prevents any downstream override.

When to lock a prompt and when not to

Prompt locking is the right choice in specific situations. It is not a default to apply to every agent.

Lock when the prompt is a compliance or safety control

Safety screeners, compliance checkers, output validators, and tone enforcers are the clearest cases for locking. These agents exist to apply a consistent standard. Variability in their prompt is a defect, not a feature. Lock them.

Lock when the agent is shared across teams

If multiple teams or workflow designers use the same agent, a locked prompt ensures they all get the same behavior. Without locking, one team might modify the prompt for their use case and inadvertently affect how other teams expect the agent to behave. Locking removes that risk.

Lock when the prompt has been reviewed and approved

If a prompt has gone through a review process (legal review, compliance sign-off, editorial approval), locking it preserves the approved version. Any change to the prompt requires going back through the Prompt Library, which is a deliberate action that creates an audit record.

Do not lock when the workflow designer needs flexibility

Some agents are general-purpose and are intentionally configured differently in different workflows. A general summarization agent might use different length constraints or output formats depending on the workflow context. Locking the prompt removes that flexibility. For general-purpose agents, an inline prompt or an unlocked library prompt is the better choice.

Do not lock when the prompt is still being iterated

If the prompt is in active development, locking it too early creates friction. Wait until the prompt is stable before locking.

Who can lock, unlock, and view locked prompts

Prompt locking is controlled by the same role-based access that governs all agent operations in DAVE. The platform ships with six default tenant roles: Admin, Create, Curate, Use, Reporting, and Financial. All roles are editable by your administrator.

Action	Admin	Create	Curate	Use	Reporting
View Prompt Library	Yes	Yes	Yes	No	No
Create and edit prompts	Yes	Yes	No	No	No
Lock prompts to agents	Yes	Yes	No	No	No
See locked prompt in Workflow Editor	Yes	Yes	Yes	No	No

Curate-role users can view the Prompt Library and see which prompt is locked to an agent in the Workflow Editor, but they cannot create prompts, edit prompts, or toggle the Lock Prompt switch. This makes Curate the appropriate role for workflow designers who should be able to build workflows using locked agents without being able to modify the agents themselves.

Use and Reporting roles cannot access the Agents section at all. The sidebar link is not visible to these roles.

Frequently asked questions

Can I lock an inline prompt instead of a library prompt?

No. The Lock Prompt toggle only appears when a Library Prompt is selected as the prompt source. Inline prompts cannot be locked. If you want to lock a prompt, it must first be saved to the Prompt Library.

What happens to existing workflow instances when I lock a prompt?

Locking a prompt affects new workflow instances going forward. Existing running instances that were already in progress when the lock was applied continue to use the configuration they started with.

Can I unlock a prompt after locking it?

Yes. Return to the agent edit form, toggle the Lock Prompt switch off, and save. The prompt selector in the Workflow Editor becomes editable again for that agent.

What if I need to update the locked prompt?

Edit the prompt in the Prompt Library. Because the agent references the library prompt, the updated prompt content takes effect for all agents that use it on their next execution. You do not need to re-lock after editing the prompt content: the lock remains in place.

Can a workflow designer see which prompt is locked?

Yes. In the Workflow Editor, the prompt field for a locked agent shows the locked prompt's name in a greyed-out state. The designer can see what prompt the agent uses, they just cannot change it.

Does prompt locking affect the AI guidelines that are prepended at runtime?

No. Prompt locking controls the agent's own system prompt. Server-wide and tenant-level AI guidelines are still prepended automatically at runtime regardless of whether the prompt is locked. The locked prompt and the injected guidelines are both applied: the guidelines come first, then the locked prompt.