Entities

Entities are structured context blobs the agent reads before acting. They are version-controlled, project-scoped, and managed with veirox entity.

The 13 entity kinds

Rules, skills, and commands can also be set platform-wide by a platform admin — visible read-only in every project alongside your own.

KindWhen to use
memoryDurable facts the agent writes and reads across sessions — services, people, recurring incidents.
knowledgeLong-lived reference docs: architecture, team conventions, how-tos.
skillReusable capability doc — describe how to do a task once, reused across sessions. Can be platform-wide.
commandCustom slash command, invoked with /name in chat. Can be platform-wide.
ruleAlways-on or path-scoped guardrail injected into the agent's prompt. Can be platform-wide.
hookPre/post-tool-use policy hook (YAML) — allow, deny, or enrich a tool call.
runbookStep-by-step operational procedures the agent follows when triggered.
issueTracked problem — status, priority, assignee.
incidentActive operational incident — severity, status, linked sessions.
postmortemIncident retrospective, linked back to the incident.
findingSecurity or quality finding — severity, status, source, CVE refs.
storyJira-style unit of work — acceptance criteria, Definition of Done.
planThe agent's ordered, status-tracked execution plan for a story.

CRUD operations

veirox entity create --kind runbook --title "Deploy checklist" --body-file ./deploy.md
veirox entity list   --kind runbook
veirox entity get    --kind runbook --ref deploy-checklist
veirox entity update --kind runbook --ref deploy-checklist --body-file ./deploy-v2.md
veirox entity delete --kind runbook --ref deploy-checklist

Versioning

Every update increments the version counter. Pass --version <n> for optimistic concurrency — the backend rejects the update if your base version is stale.

See also