Specs and AI Prompts
Capture durable specifications and reusable AI prompt templates so your agent always has the right context.
Two Developer Mode features give an AI agent the context it needs to do good work: Specs and AI Prompts. Both are available once Developer Mode is enabled on a project.
Specs
A spec is a durable specification — an API contract, a data model, a design decision, a set of acceptance criteria. Specs are owned at the project level and can be linked to multiple milestones, so a single source of truth is reused wherever it's relevant.
Specs carry version history: every change is recorded, so you can see how a specification evolved and compare versions.
There are two flavors:
- Global specs — project-wide references, linkable to any number of milestones.
- Acceptance criteria — a spec tied directly to one milestone, describing what "done" means for it.
When an agent fetches a task, it can also pull the milestone's specs, so it implements against the agreed contract instead of guessing.
AI Prompts
AI Prompts are reusable prompt templates stored in the project. They standardize how you ask an agent to do recurring work — planning a task, writing a spec, reviewing a change — so everyone gets consistent results.
Briefboard provides a set of default prompts, and you can:
- Customize a default by overriding it with your own version.
- Create custom prompts specific to your project.
- Restore a default you've hidden or changed.
Prompts pair naturally with specs: a planning prompt can reference the milestone's specs, turning agreed criteria into an actionable implementation brief.
Putting it together
A healthy loop looks like this:
- Write a spec describing what you want and link it to the milestone.
- Use an AI Prompt to have the agent plan the task from that spec.
- Let the agent implement over MCP, checking against the acceptance criteria.
The specs keep intent explicit, the prompts keep the process repeatable, and MCP keeps the agent working from live data.