After Your First Ralph Workflow Run¶
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.
If Ralph Workflow did its job, the question after your first run is not whether the agent sounded smart.
It is this:
does the implementation hold up?
This page is the shortest path from a private first run to a useful public next step on Codeberg.
Primary repo: https://codeberg.org/RalphWorkflow/Ralph-Workflow
GitHub mirror only if that is where you already follow projects: https://github.com/Ralph-Workflow/Ralph-Workflow
Use Codeberg for the real project relationship: stars, watches, issues, and contribution history.
The only branch that matters¶
Treat the first run as a fork with only two healthy outcomes:
It earned trust → star or watch Ralph Workflow on Codeberg.
It exposed friction → open the matching first-run or docs/proof issue form on Codeberg.
If you do neither, the adoption signal stays private and the project learns nothing.
The two-minute scorecard¶
After the run finishes, check these in order:
Does the diff match the task?
Did the software actually do the requested thing?
Did the checks actually run, including integration checks where they matter?
Are the changes reviewable in one sitting?
Are open questions called out clearly?
does the implementation hold up?
If the answer to most of these is yes, Ralph Workflow is doing useful work.
If the answer is no, the failure is still useful if you can say why.
What to do next on Codeberg¶
If the run was genuinely promising¶
Use the primary repo on Codeberg:
Star the repo — https://codeberg.org/RalphWorkflow/Ralph-Workflow
Watch the repo — https://codeberg.org/RalphWorkflow/Ralph-Workflow
Keep the repo handy for the next task — https://codeberg.org/RalphWorkflow/Ralph-Workflow
That is the cleanest public signal that Ralph Workflow earned a place in your toolset.
If the run exposed friction¶
Report it on Codeberg so the fix lands on the primary repo:
First-run friction form — choose that form at https://codeberg.org/RalphWorkflow/Ralph-Workflow/issues/new
Docs / proof gap form — choose that form at https://codeberg.org/RalphWorkflow/Ralph-Workflow/issues/new
Best bug reports say:
what real task you tried to hand off
where the friction showed up
what you expected instead
what would have made you more likely to trust or keep using Ralph Workflow
What counts as a successful first run¶
A successful first run does not need to be perfect.
It only needs to prove something real:
the task was scoped well enough to hand off
the workflow came back with real executable proof
you can tell why you would or would not merge it
That is already much better than a transcript that only sounds finished.
When to give it a harder second task¶
Move to a bigger second task only if the first run gave you:
working behavior you can verify
real checks
a clean re-entry point the next morning
enough trust that you would review another result
If not, do not jump to a bigger task yet.
Instead:
tighten the spec
choose a more bounded task
report the friction on Codeberg
If you are still unsure¶
Use these in order:
The goal is simple: convert your first run into one honest public action on Codeberg — either a star/watch because it worked, or a useful issue because it did not.