The admin tools

The nine actions the admin agent can call. Every call runs as the conversing user and needs the ContextWorks Admin or ContextWorks Domain Manager permission set. Reads return aggregates and configuration, never record data — record data stays on the runtime actions.

Seven of the nine are administration-only. Get Catalog and Get Skill are the runtime actions, registered on the administration topic as well.

Get Catalog

The same action agents call at runtime — the index, an entity’s full contract, scoped by domain — so you can read exactly what an agent receives. Inputs: Invocable actions.

Get Health

How ContextWorks is doing. One area per call.

InputNotes
areatrends (default — this 7 days vs the prior 7, by domain) · usage · quality (entities only) · issues · writes
domainApiNameOptional: only traffic through one domain
libraryOptional: entity, brief, skill, or catalog
artifactNameOptional: one artifact
daysDefault 7, max 90
formatmarkdown (default) · json

Answers are org-wide, like the Observe tab — managers read the same numbers admins do.

Get Change Log

The configuration ledger: who created, activated, rolled back, deactivated, or deleted what, when, with which comment. Managers see the domains they manage.

InputNotes
libraryOptional: entity, brief, skill, or domain
artifactNameOptional: one artifact
daysDefault 30, max 365
maxRowsDefault 50, max 200
formatmarkdown (default) · json

Get Full Config

The configuration graph, scoped by domain, object, or entity list. Two shapes. analysis is the default: activation state, what each domain carries, and effective writability. export returns the round-trip bundle. To read the whole configuration with no scope, pass unscoped=true.

Save Draft

Compose or update a draft. A draft serves no one until activated.

libraryTakes
brief (default)Type, label, description, content, refusal policy, customer-safe, synonyms, review-by, home domain
skillLabel, description, procedure, output shape, surface, synonyms, home domain
entityDescription, instructions, synonyms — text only, existing entities only

A domain manager drafts into domains they manage, and a new artifact has to name a home domain they manage. Every brief the agent drafts carries a stamp naming you: Drafted by an agent acting as ….

On Agentforce, Save Draft is the one packaged action that asks you to confirm before it runs.

Activate Draft

Publishes a draft as the next served version.

  1. The agent calls without confirm. The preview shows the current version, the changed parts, which domains carry it, the exact content that will serve, and a digest.
  2. Approve the preview.
  3. The agent calls again with confirm=true, the digest, and a comment — the comment lands in the Change Log.

Change the draft after the preview and the digest goes stale, so ContextWorks refuses the commit. Preview again.

Update Domain Contents

Adds or removes an entity, brief, or skill from a domain, with the same preview-then-confirm flow. The preview shows what the domain carries before and after, and whether the change reaches users immediately — it does when the artifact is active. You need rights to change that domain’s roster.

For a brief, delivery picks how surely the body arrives: offered (on request through Get Briefs), catalog (the body also rides the catalog response, once per discovery call), or injected (the body rides every payload the domain serves). → Delivery

Deactivate

Withdraws an artifact from service immediately. The preview names what stops serving and where; confirm=true pulls it. Versions are kept — re-activating restores service.

Get Skill

Fetches one procedure by API name; the agent then follows it. The same action agents call at runtime, so it reaches both the seven packaged admin procedures and the runtime skills you hold. See the skills.

What’s next