agentos

Resolve, inspect, apply, verify, review or improve an AgentOS Agent composition. Use when selecting instructions, Skills, harness-native settings or capability requirements for a persistent Mate or bounded Assignment; loading an external composer or origin; reconciling desired and observed setup; collecting a composition debrief; or routing evidence into a separate improvement.

Select, configure, launch, inspect, resume, or change a verified AgentOS coding harness. Use before a First or Second Mate chooses a Crewmate harness, model or reasoning effort; before launching Pi or Codex through Herdr; or when harness authentication, flags, defaults, quota or recovery affect dispatch.

Inspect, create, and operate AgentOS runtime pods, Kustomize overlays, Herdr sessions, Mise toolchains, agent worktrees, terminals, PVCs, Kubernetes health, attach paths and runtime recovery. Use for normal First- or Second-Mate Kubernetes operations, child-Agent workload creation, tool resolution, stuck-agent debugging, remote terminal access, pod or session recovery, and readiness diagnosis.

Reconcile and supervise an AgentOS Mate's direct reports across PostgreSQL, Kubernetes, Herdr, PVCs, and Git. Use at every First- or Second-Mate session start, after restart or compaction, while delegated work is active, when Inbox or runtime state changes, and for blocked, stale, stuck, failed, interrupted, or recovering agents.

Inspect an environment and interactively establish or reconcile a persistent AgentOS First Mate in Kubernetes. Use only for first installation, a partial bootstrap, local-clone onboarding, cluster selection, or handoff from a temporary local agent to the cluster First Mate.

Charter, provision, route to, recover, change, and retire persistent AgentOS Second Mates. Use only from First Mate before any Second-Mate lifecycle operation, scope decision, charter handoff, runtime recovery, work transfer, configuration change, or retirement.

Inspect, provision, migrate, verify, and recover the AgentOS PostgreSQL database using released SQL assets. Use for database topology, schema versions, Fleet coordination, Tasks, Assignments, Inbox, Captain state, external-event reconciliation, roles, grants, RLS, Functions, Triggers, migration failures, or PostgreSQL bootstrap and recovery.

Route and deliver project-specific work through AgentOS Tasks, Assignments, Second Mates, ship Crewmates, and scout Crewmates. Use before accepting, planning, investigating, coding, auditing, assigning, briefing, spawning, merging, handing off, completing, or retiring delegated project work.

Evaluate an Agent organization through the public AgentOS benchmark. Use when measuring Quickstart, Fleet delivery, composition integrity, recovery, human attention, efficiency, robustness, safety or accountability; running conformance, live or offline evaluations; or producing and validating a sanitized evidence bundle.

Turn a frozen AgentOS evaluation into the smallest reviewed, regression-tested improvement. Use after benchmark or operational evidence exposes wrong tool use, repeated trial and error, unclear instructions, stale Skill or composition guidance, runtime failure, an authority violation or another reproducible Fleet weakness.

Install, configure, inspect, recover, rotate, or retire the optional AgentOS Fleet AI Gateway for pooled Codex subscriptions and an explicit OpenAI API-key fallback. Use when a First or Second Mate considers shared AI capacity, server-owned provider OAuth, native Pi or Codex routing through the gateway, quota-aware account routing, gateway client Secrets, or gateway 401/429 failures.

Establish, inspect, rotate, revoke, and recover model-provider or source-provider authentication for AgentOS agents. Use for Pi Codex-subscription browser login, GitHub personal or App identity, device-code recovery, API-key or third-party fallback, provider changes, expired credentials, quota identity changes, and authentication verification.

Develop, review, test, dogfood, or deliver changes to AgentOS itself from a local checkout or a running Fleet. Use when an Agent is asked to modify the AgentOS repository, choose an upstream versus organization-specific delivery path, prepare an AgentOS pull request, or run an unreleased AgentOS revision in First- or Second-Mate pods.

Resolve and safely change an AgentOS project's local and remote lifecycle. Use before registering, cloning, creating, initializing, changing delivery or autonomy posture, removing, or reconciling a project checkout or provider repository.

Preserve unresolved Captain choices discovered by investigations or reviews and route their answers durably. Use before completing a Scout or structured review, when linking dependent Tasks to a Captain choice, and when recording or resolving the Captain's exact answer.

Establish falsifiable causality for a reported software defect. Use before a First or Second Mate briefs a bug Scout, when evaluating a diagnostic report, and before authorizing implementation from that report.

Select, prepare, operate, and retire ArtifactFS-backed Scout workspaces for fast read-heavy access to large or multiple Git repositories. Use when a First or Second Mate is delegating a scout investigation whose clone startup or cross-repository breadth justifies an isolated FUSE-enabled Crewmate profile; never use it for persistent Mates or ship work by default.

Build, test, and publish OCI images safely from an exact Git revision inside or outside an AgentOS Fleet. Use when a First or Second Mate needs an unreleased image, must choose BuildKit or Buildah for an in-cluster build, or needs to make an immutable image digest available to Kubernetes without a host runtime socket.

Select, establish, inspect, or retire an OCI registry for AgentOS and other Fleet-built software. Use when images need a registry, when a Fleet may deploy zot, when kubelet pull reachability or registry trust must be designed, or when choosing between a disposable development registry and a durable private organization registry.