How a run works
1
Find undocumented components
The auditor scans your synced Salesforce catalog for components with zero doc tags — nothing in your docs claims to cover them. It respects your coverage eligibility settings, so components you’ve excluded (for example, managed-package namespaces) are never proposed. See Coverage.
2
Rank by importance and recency
Candidates are ranked so the components that matter most — heavily referenced, recently changed — come first.
3
Check your docs corpus first
Before drafting anything, the auditor searches your existing documentation. If a component is already meaningfully mentioned somewhere, it’s skipped — no duplicate pages.
4
Require real evidence
The auditor only writes about what it can actually see. Components with hidden or empty source are skipped rather than guessed at, and skips are remembered for 30 days so the same dead ends aren’t retried every run.
5
Draft and file
For each surviving candidate, the auditor drafts a doc per configured doc type, grounded in the component’s real metadata and shaped by your templates and style. The batch is filed into the Inbox as new-doc proposals.
This is the same rule every LivingContext agent follows: grounded or it doesn’t ship. A proposal without evidence is a skip, not a guess.
Configure it
Configuration lives at Settings → Doc suggestions:
A status line shows the cadence and the last run’s outcome — how many undocumented components were scanned and how many docs were proposed. Scheduled runs ride the shared agent sweep, which checks every 15 minutes whether a run is due — enabling the schedule starts the first audit within one tick. See Agents overview.
Reviewing proposals
New-doc proposals appear in the Inbox alongside commit-driven updates, grouped as suggestions. For each one you can read the full draft, see why the component was proposed, and accept or dismiss it. When you accept and ship a proposal, the new doc is automatically tagged to the Salesforce component that prompted it — so the component immediately counts as documented in Coverage, and future org changes to it will trigger freshness checks.What it won’t do
- It never publishes directly — every draft goes through Inbox review and lands as a normal change in your Git repository.
- It never proposes components excluded by your eligibility settings.
- It never fabricates content for components whose source it can’t read.