The only way to know whether a writing profile is correctly configured is to see what it generates. You can read through every setting, adjust the tone descriptors, refine the audience framing — and still have no reliable way to predict the output until it exists. The Profile Sandbox is how you get that output without it costing you anything that matters.
What the Profile Sandbox Is
The Profile Sandbox is an isolated workspace for testing a writing profile. It lives inside the profile editor, not inside the kanban pipeline. Runs you do there do not appear in post history. Nothing generated in the sandbox enters your content pipeline unless you explicitly choose to act on it.
This matters because profile tuning and content production are two different jobs. The sandbox keeps them structurally separate — you work on your profile in one place, you produce content in another, and the two do not interfere with each other.
What It Does
Without the sandbox, every profile test would require burning a real idea card and consuming one of your daily drafts. You would make a change to the voice settings, run a draft, decide the output still isn't right, make another change, and repeat — each iteration costing you something real.
The sandbox breaks that loop. You can test a profile change, read the output, decide it needs adjustment, and run again. The cost per run is between $0.02 and $0.04, metered to your account usage history, and it does not count against your daily draft limit. Profile tuning stays self-contained.
How a Sandbox Run Works
To start a run, you provide a topic or working title — this is the only required input. Optionally, you can add Notes to give the run more context, attach References as URLs or files, and select which profile to test if you have more than one. The selector defaults to your currently active profile.
The sandbox runs Pass 2 only. Research and Pass 1 (outline generation) are skipped. The draft is generated directly from your topic, notes, and references against the current profile state. This means you get output faster, but it also means the draft is built without the research layer that a full pipeline run would include — the sandbox is testing the profile, not producing a finished piece.
Run time is roughly equivalent to a standard Pass 2 draft generation. The output appears in the sandbox view alongside the current profile state, so you can read the draft and the profile settings that produced it at the same time.
The Approval Gate and Version History
Profile edits made inside the sandbox view do not affect your live profile. You can adjust settings, observe how the draft would have read with those changes applied, and iterate freely — nothing is live until you trigger Approve Profile Changes.
That action copies the sandbox profile state to the live profile and creates a new version record with an approval timestamp. From that point forward, new pipeline runs use the updated profile. Posts that were already generated are unaffected — they were produced against whichever profile version was live at the time.
Every profile change creates a new version automatically, whether the change originates in the sandbox, the guided editor, or advanced markdown mode. There is no explicit save step. Version history stores full profile snapshots — not diffs — along with timestamps, optional labels (for example, "after sandbox run 3"), references to which sandbox run produced a version, and rollback references.
Sandbox run history is preserved after approval. Runs are not deleted when you promote a profile state.
Rolling back to a previous version creates a new version with the content of the target version, timestamped at the moment of rollback. History is always additive. No version is ever overwritten or auto-deleted.
Limitations and Expected Use
Sandbox output is not publishable directly. If a run produces a draft worth keeping, you need to create a real idea card and run the full pipeline. The sandbox exists to tune the profile, not to produce content.
Most users reach a satisfying profile state in two to five runs. Users with distinctive voices or specific content types — formats that behave differently from general prose, strong stylistic constraints, highly specific audience assumptions — may need more iterations before the output reliably matches what they're looking for. That's expected and the cost structure is designed for it.
After approval, prior posts are not retroactively affected. Each post in your history was generated against the profile version that was live when it ran. A profile update changes what comes next, not what already exists.
To open the sandbox, go to the writing profile editor and select the sandbox tab. If you've just completed the profile setup wizard on a new account, Riterly routes you there automatically before your first real pipeline run — that sequencing is intentional. The sandbox is the right place to confirm your profile is behaving the way you expect before idea cards start moving through the pipeline.