Product guides

Product building blocks

Each primitive handles a different part of the run.

PrimitiveWhat it holdsWhat it changes
Run workspacesMutable files, tools, browser state, and intermediate results for one runGives the agent a place to do the work
Documents and artifactsAccepted deliverables and retained previewsPreserves the canonical result beyond the workspace
Automations and eventsTriggers, conditions, actions, and activity historyStarts work at the right time
IntegrationsGoverned access to business and developer systemsLets the run operate where the work lives
KnowledgeApproved facts, policies, examples, and organization contextGives the run the right information
SkillsVersioned procedures and checks for repeatable workGives the run a tested method
MemoryScoped guidance from accepted work and correctionsImproves future choices without rewriting every prompt
Identity and permissionsRun owner, tool allowlists, network boundaries, approval rightsDefines what the run may see, change, and approve
ReviewEvidence, questions, changes, and approval gatesKeeps consequential decisions accountable

The prompt describes today's job. Knowledge supplies approved facts, a Skill supplies the method, and Memory offers scoped guidance from earlier reviewed work. The named system of record and the accepted artifacts keep current truth separate from working copies.

Which layer a correction belongs in

Most operating mistakes come from putting a correction in the wrong layer. A policy pasted into a prompt has to be pasted again. A one-time exception promoted to Knowledge becomes a rule that quietly breaks the next run.

The correction isIt belongs in
A fact or a policyKnowledge
A change to the procedureThe Skill
A one-time exceptionThe run, and nowhere else
A missing checkA validator
The wrong tool for a stepA tool rule or routing guidance