Spec-Driven AI Agent: Why the Spec Matters More Than the Prompt¶
Ralph Workflow is a free and open-source AI agent orchestrator built around a simple core loop inspired by the original Ralph loop. That simple core composes into a stronger workflow system for serious repo work, and the default workflow is already strong enough to start with before you customize anything.
Ralph Workflow is a free and open-source spec-driven AI agent workflow for developers who want results they can actually review instead of transcripts they have to decode.
If an agent keeps saying it is done before the work actually holds up, the problem is often not raw model capability. The problem is the absence of a real spec.
What “spec-driven” actually means¶
A spec-driven AI agent does not start from vague intent alone. It starts from a written task that makes four things explicit:
what should change
what should stay unchanged
what done looks like
what checks prove the work holds up
That is the difference between hoping the agent converges and giving the run a finish line.
Why specs beat prompts on substantial work¶
Prompts tell the agent what to do. Specs tell the agent what success looks like.
That matters most when the task is too big to babysit and too risky to trust blindly. Without a spec, “done” becomes the agent’s opinion. With a spec, “done” becomes something you can verify.
The workflow Ralph Workflow is built for¶
Ralph Workflow is built around a spec-first loop:
write the task in
PROMPT.mdrun planning, implementation, verification, and review
come back to a real diff, checks, artifacts, and open questions
decide whether you would merge it
That is what makes it different from a normal AI coding chat. The point is not to produce a plausible answer. The point is to run a real software workflow with executable proof and verification.
Who this is for¶
Ralph Workflow is for developers and technical teams who already use coding agents on their own machine and want a better way to hand off meaningful work overnight.
Good fit:
bounded features
refactors with clear acceptance criteria
test expansion
cleanup work with obvious verification
Bad fit:
vague exploration
risky production surgery with no harness
tasks where nobody agrees what success looks like
Best first evaluation path¶
Inspect the primary Codeberg repo first: https://codeberg.org/RalphWorkflow/Ralph-Workflow
Use Getting Started to run one real task
Use Choose Your First Ralph Workflow Task and How to Review AI Coding Output Before You Merge to judge the result honestly
Turn that result into one public next step with After Your First Ralph Workflow Run
Use GitHub only as the mirror if that is where you already track projects: https://github.com/Ralph-Workflow/Ralph-Workflow
Best next step on Codeberg if you want a spec-first overnight workflow¶
Do not leave the evaluation private.
Use Codeberg as the main public home:
Inspect the source on Codeberg: https://codeberg.org/RalphWorkflow/Ralph-Workflow
Star or watch on Codeberg if the workflow earns trust: https://codeberg.org/RalphWorkflow/Ralph-Workflow
Report first-run friction or spec/proof gaps on Codeberg: https://codeberg.org/RalphWorkflow/Ralph-Workflow/issues/new
Use GitHub only as the mirror: https://github.com/Ralph-Workflow/Ralph-Workflow
That keeps the trust signal and the feedback loop on the primary repo instead of splitting them across surfaces.
Why try it now¶
Because Ralph Workflow is free and open source, runs with the agents you already use on your own machine, and lets you test a spec-driven overnight workflow on one real backlog task tonight.
Run one real task, judge the result honestly, and then take exactly one public action on Codeberg:
promising run: star or watch the repo
shaky run: open the right issue on Codeberg