Two audits in one click: structure, then meaning

Stage one: structural checks

Click “⚙ Audit + AI review” and the first reply is instant, because it is not AI. Runic parses the process and walks it: a decision with fewer than two branches (“This decision only covers ‘Yes’ — what happens otherwise?”), a branch with nothing in it, a step that sits after a .go or after the End and can never run. Two softer hints join them: a step longer than 45 characters that reads like two actions joined by “and”, “&” or “, then”, and a process of eight or more steps with no .if and no .fail — “does every run really go perfectly?”

Each finding shows its line as a chip (“L12”); click it and the editor jumps to that line and flashes it. Findings are sorted by line and marked warn or hint. A clean pass says so: “decisions have branches, every step is reachable.”

Stage two: the analyst

Structure is half an audit. The second stage sends the document to the chat route with a fixed brief: review it like a senior business analyst and say what is missing or risky — undocumented failure paths, unclear roles or permissions, data collected but never validated, compliance or domain gaps — citing line numbers, concrete and brief, no praise, in the language the document is written in. The recent conversation and the archive summaries travel with it, so the analyst knows what you have already discussed.

Because the brief asks about compliance, the router allows web search on this call. When the analyst actually looked something up, the reply is tagged “web research” and is asked to end with a Sources line.

It also runs by itself

Every time the AI writes into a process — an applied proposal, a Create with AI draft, an imported flow, an accepted inline edit — the structural pass runs automatically. If it finds anything, a message appears: “Structural check on ‘<document>’ — 2 issues in what was just written (run ⚙ Audit + AI review for the analyst pass)”. If it is clean, nothing is posted. The automatic check reports; it never edits.

The same checks guard the AI itself

The hard findings — one-branch decision, empty branch, unreachable step — plus parse errors are the score Runic uses to self-repair AI output before you see it. A document that scores above zero goes back to the model once with the validator report; the fix is kept only if it scores better. So by the time you press Audit, the obvious structural faults have usually been caught upstream, and what remains is the analyst’s job.

Limits

Audit exists for processes; the button is hidden on sequence diagrams. Findings are conversation starters, not errors — syntax problems are the parser’s job and show in the editor. The analyst pass counts as one chat message against the 60-an-hour limit. The document-level hint about no failure path has no single line, so it shows a dot instead of a line chip.

How to use it

1. 1
   Open a process

   The Audit button only appears for process documents.
2. 2
   Click “⚙ Audit + AI review”

   It sits at the bottom of the chat panel, in the row just above the message box, next to 📎 Import flow. The tooltip reads “Instant checks plus an AI analyst pass”.
3. 3
   Work through the structural list

   The first reply lists findings with L-number chips. Click one to jump to that line in the editor.
4. 4
   Read the analyst

   The second reply arrives after “Reading the process… Thinking it through… Writing…”. It cites line numbers and may end with Sources.
5. 5
   Fix from the chat

   Reply with what to change — “add the missing No branch on line 12” — and apply the resulting diff.

Commands & shortcuts

Tips

• Run the audit before sharing a process; a clean structural pass and a short analyst list read well in a review.
• Answer the analyst in the same thread. “Add a .fail for the payment timeout you mentioned” turns a finding into a diff.
• The automatic check after an applied change is the same stage one. If it is silent, the structure held.
• The structural pass is also exposed to Claude over MCP as audit_document, one document per call.

Questions