Agent view reference

These commands report launch previews, current manager observations, and retained team receipts.

Persona previewsCopy Anchor Link

spawn --file/-f <persona-file> selects a persona, not a team manifest. It cannot be combined with a positional persona or the persisted internal --config spelling. A bare filename is still a file, including when it has no extension. Managed paths remain relative to the manager's workspace and are read on that execution host. No file upload is implied. Direct foreground execution reads its local file. A supported Mesh-manifest shape returns a schema refusal with the literal cotal team deploy <file> remedy, before agent allocation.

spawn --dry-run never refreshes the catalog, repairs stale registrations, imports connector code, signs in to the broker, redeems enrollment, provisions or allocates an agent. Where the launch reads whether the recorded brokers answer to choose its space, as in a folder whose registered space may be stopped, the preview reads the same without signing in, so it names the space the launch uses. The managed preview shows the request and exits 0 with completeness not-checked, because the manager resolves the persona, harness and working directory on its own host at launch. It does not substitute the caller's files. A named --runtime is checked on this machine only, through the lookup the launch uses: result.requested.runtime names it, and result.runtimeCheck gives the host it was checked on, whether it is available there, and the error when it is not. A runtime this machine cannot run is a RUNTIME_UNAVAILABLE warning, because a manager on another host checks its own. --runtime auto is refused as the launch refuses it. With --foreground, the preview resolves the local persona, harness precedence and physical working directory. Binary, login, model-policy and live authority checks are labeled not performed. Both previews say where the persona came from: --file, the command line, a preference or the default. A harness that names no connector this workstation knows is a HARNESS_UNKNOWN warning, or the PREFERENCE_INVALID warning config gives when only COTAL_DEFAULT_AGENT chose it. In a folder whose .cotal/ names no space, as after space forget, the preview keeps Mesh: main, says under it that the held default was assumed, and its JSON target names the folder as root and the space as assumed.

spawn --foreground --enrollment-file <private-file> --space <name> uses the existing one-seat foreground enrollment path. The file is opened with protected ownership/mode and bounded input checks; its URL is not printed, copied into the child environment or persisted by this input adapter. It does not supply a managed enrollment authority path. Dry-run does not read or redeem the enrollment file. On Windows the file must be a regular file; its owner, mode and links are not checked.

Managed launch viewCopy Anchor Link

spawn [<persona>] --on <instance> launches through that manager and attaches after its native readiness result, when that manager's runtime streams a terminal. The attachment is pinned to the attributed accepting manager and the allocated identity/incarnation, not a second display-name guess. --detach/-d returns a finite JSON/plain-capable receipt after the same readiness result. While it waits, it prints the launch it asked for, with the space as a field, and one line that says the manager answers when it refuses the launch, or when the agent joins or exits, within about 30 seconds. The receipt or refusal that follows names the mesh in its Mesh: heading. A launch the manager reports failed, such as an agent whose process exits at once, fails OPERATION_FAILED, exit 5, with the manager's reason. The terminal receipt shows the full lifecycle, original request, accepted goal where reported, role and model request separately from any provider acceptance claim. When the host granted the agent other read or post channels than it asked for, as a host that narrows a request to your grant does, the receipt adds a Granted line with the channels it got and the ones it asked for, and --json carries both as acl.requested and acl.effective. Its Next section names the actual agent, space and accepting manager in attach/status commands when their tokens are safe shell words; it otherwise retains placeholders beside the full identity fields.

A manager on a runtime that streams no terminal, which is tmux, cmux, orca and herdr, answers with that fact and the launch stops there: no attach is asked for, nothing is partial, and the command exits 0 with the agent running. Its Runtime row reads <runtime> · no terminal stream, its View row reads not attached; the seat runs in the manager's <runtime> session, and its Next gives cotal watch and cotal send -t <name> with one line saying attach is not available there. --json carries the same answer as result.streamsTerminal, beside the result.runtime that names the backend. A manager from before 0.104.2 reports neither, so streamsTerminal is null and the launch asks for an attach. When that manager refuses it because the runtime streams no terminal, the launch ends there too, with one line that names the runtime, and exits 0. Such a manager does not mark that refusal, so the CLI accepts it only for a runtime whose provider, registered or installed with this CLI, declares no terminal. Any other failed attach is partial, exit 4, with the manager's reason as Native detail and a Next that gives cotal ps and cotal attach with the space and manager instance. Without usable input/output terminals, an attaching --on launch refuses before allocation and names -d or a spawn in this terminal. A shell tool's inherited parent PTY is not enough.

A launch whose agent application waits for you before it joins is ready once its process runs. Claude with --opt resume= and no session id opens its session picker, so the CLI says the agent is running on its manager and host and waiting for you at Claude's session picker. In a terminal it attaches to the agent at once, -d included; otherwise it prints the attach command. When the readiness window passes and the manager reports the agent's process still running, the CLI says the agent has not joined the mesh yet, attaches when the spawn was not detached, and otherwise prints the attach command. Both exit 0, and --json reports them as managed-launch-running with waitingFor naming the screen or null. The outcome stays unknown when the manager cannot say whether the process runs.

A managed launch prints its Managed launch plan on stdout before its receipt, so cotal spawn <persona> -d > out.txt keeps both, and a refusal prints on stderr. With --json, stdout holds only the envelope and the plan stays on stderr. The manager that refuses a launch writes one line to its own log naming the original request, the persona and the reason, so cotal service logs <definition> on its host shows why a spawn failed after the terminal is gone. An unpinned launch that no manager received has no such line, because no manager answered it.

--harness/-H, --model/-m, --name/-n, --prompt/-p and --cwd/-C retain their declared meanings. The legacy --agent spelling remains accepted; using both harness spellings refuses. A --file <persona-file> launch uses this route too unless --foreground is given. A manifest is deployed with cotal team deploy <file>, not spawn.

--foreground runs the native application directly. It does not support a structured receipt, manager pin or manager control credential. Native exit or terminal loss may end that process; no detachable-view or reboot-survival claim applies. Protected internal handoff/enrollment children keep their existing direct execution and native authorization checks.

If the agent joins but attachment fails, the result is partial with its original identity and manager. No second launch occurs. A view exit sends no stop request. This route does not offer no-manager startup choices or imperative planning.

Named agent statusCopy Anchor Link

cotal status <agent> reads the authorized manager census and separates Identity, Execution, Mesh and Last reported activity. JSON has identity, execution, mesh and activity fields; missing optional observations are null with source annotations. --on <manager-instance> restricts the read. It takes no --all or --components. cotal status manager, or an answering manager's instance id, names the manager itself, which runs no row of its own: it refuses TARGET_NOT_FOUND like any unknown name, and its Next says the manager is answering and points to cotal status and cotal endpoint describe manager. A requested model is not provider-confirmed. Connector-reported provider information is labeled exact or name-matched under the same proof. A new manager row marks a roster observation exact only when its principal and lifecycle UID match the managed seat and its presence view is current. Older or mismatched rows leave exact presence and activity null and label any name-matched values separately. A verified condition and its reported start appear with activity; an unverified condition stays in a separate name-match field. A waiting observation is not a measured blocked duration, and a work event is not task completion. Attach and stop hints retain the answering manager's instance when it is a valid native id.

A same-name match on several managers refuses as ambiguous. An unanswered manager leaves the search partial; no matching row in that partial read is not proof the agent is gone. The command starts, stops and attaches nothing.

Recorded team viewsCopy Anchor Link

Recorded launch evidenceCopy Anchor Link

team (t) reads fresh launch summaries, additive ownership ledgers and their root-local attempt records at the selected local root. A team read or --dry-run preview that completes exits 0. Its JSON marks completeness as partial, and its report has a line saying what it does not establish. A report names each record's kind, recorded state, next action and auth mode in words, such as boot returned and static auth. The JSON result keeps the recorded values, such as broker-boot-returned and auth. A deployment ID that no record in the root names is refused with TARGET_NOT_FOUND, which names the ID, by team show and by team remove; team show adds that an ID is 16 hex characters when the value has another shape. When native up -f returns from its broker start, it writes a creation-only fresh summary with its full run ID, initial broker PID, declared agent/channel names and the initial control-plane, auth-service and delivery-responder results. Older summaries without a broker PID, auth-mode pin or responder observation remain readable; team show labels their absent facts not recorded and says the record proves no current ownership. A PID or a responder bound when boot returned is not current process or listener ownership proof. A private create-only fresh boot attempt is written and fsynced before the launch spec and broker start. A returned summary is also published create-only and fsynced; an intent without a matching auth-mode-pinned summary showing a returned control plane, auth service and (for auth mode) bound delivery responder stays unsettled and blocks a later fresh boot. team show <id> and the team inventory expose the original intent without claiming whether its broker started. team show displays up to 100 declared names from a summary and retains the complete list in JSON; none are claimed to have joined. A missing or refused summary after boot is not a reason to boot again. A summary records the absolute folder up -f ran in. After that folder is moved or copied, the summary still reads but names the old path. It settles nothing in the new folder and blocks no up or down there, and team show <id> refuses it with CONFLICT, naming the folder it was recorded at. The inventory's completeness stays partial because older fresh boots may have no summary. The terminal lists up to 100 records; JSON retains all readable records. Its Next column says inspect the stack for a fresh record or review the native scoped teardown plan for an additive ledger; neither row grants permission to stop anything. A fresh summary is not an additive ownership ledger or evidence that an agent joined, and it does not establish current process health. For a fresh summary, team show <id> names the recorded root and offers cotal down --space <space> --dry-run, the preview of the guarded whole-stack stop, with the recorded space filled in. team remove <id> refuses a fresh boot ID with CONFLICT and offers the same stop. The summary is not authority to run it.

Additive deployment recordsCopy Anchor Link

For an additive record, team show prints up to 100 full recorded agent identities and lifecycle UIDs where the ledger has them, with the full validated ledger in JSON, without claiming those agents are still live. Its follow-up uses down -f <file> --run <id> --dry-run from the recorded root. The file is a required argument, but the full run ID selects the ledger; an edited manifest is not new teardown authority. Review the native plan before a separate apply. Listing ownership does not confer permission to stop anything. One guarded team deploy <file> route accepts a channel-only manifest whose declared cards are all absent and explicitly set deliveryClass: live. It refuses agents, runtime requests, defaults, broker bind changes, access edges, existing cards, a changed re-apply, registered broker lists and unresolved prior attempts before an effect. A card that does not set deliveryClass: live is refused with UNSUPPORTED. The refusal names the cards, and its Next says to set deliveryClass: live on each card or to create a durable channel with cotal channel set <name> --delivery durable. team deploy --dry-run warns LIVE_DELIVERY_REQUIRED for the same cards, so the preview says the apply would refuse. It selects the exact registered root, single broker address, auth mode and TLS requirement, then holds the local maintenance lock. The original user bearer and its deployer and writer views must name the same owner and actor. A refused view stops before the intent; no static or broader caller is substituted. A scoped writer uses one non-reconnecting broker connection and KV create-only on the existing bucket, never the legacy merge-and-put path. The private intent is published and synced before any create. Every channel has a synced submitted step before the native request and a synced accepted revision after its returned acknowledgement. The returned receipt matches every declared channel and both steps. A missing acknowledgement or receipt leaves the attempt unsettled; no create is retried. If closing the writer connection fails after the receipt is confirmed, the result keeps the confirmed revisions and warns about the close. A returned receipt proves only historical created revisions, not the current cards, readers or removal authority. team show <id> presents one row per returned channel; incomplete attempts show their submitted/accepted steps. The JSON result retains both forms of evidence. Team inventory and status show returned versus unsettled state. Additive broker addresses in reports omit URL user info, queries and fragments; the protected intent keeps the exact connection target. Invalid or orphan steps stay visible and blocking. Guarded channel-only card removal uses the returned create revisions as described below. Completion reads those receipts locally; it does not query the broker or guess ownership from an edited manifest. Channel apply emits one finite rich, JSON or plain result.

Agent deployment applyCopy Anchor Link

The agent route accepts one to 100 all-new agents on a registered static-auth or open mesh, optionally with up to 100 new channel cards that explicitly use live delivery. Defaults, bind changes and stale waivers refuse. User-auth agent apply refuses before inline persistence. The route selects a manager through an attributed status read, then pins credential issuance and every subsequent request to that full instance. It requires native execution and a matching runtime; it never starts a replacement manager. --runtime overrides the manifest's requested runtime but does not install or switch a provider.

A mixed deployment also requires every requested application to appear available in that manager's connector inventory. This is a startup observation, not proof that a later launch will succeed. A separate deployer reader checks the requested channel keys, and a separately scoped writer uses one non-reconnecting connection for native KV create-only. Manager credentials are not used as channel authority. Existing cards refuse; no patch, adoption or legacy merge-and-put occurs.

One private intent records all channels, ordered resolved agent hashes and the manager before effects. Channels are created first, with submitted and accepted revision steps synced before continuing. Each agent's original request is synced before submission, its native allocation before goal following, and its matched joined result before the next agent. A joined result must match the accepted name, principal, lifecycle, harness and runtime. Unreadable or unknown phases remain unsettled and are not replayed. The full receipt is saved only when both channel and agent sequences match the original intent.

A later failure retains earlier channel creates, allocations and joins without rollback. When it created anything, its Next adds cotal team remove <id> --dry-run. Writer-close uncertainty is reported separately from returned native receipts. The finite JSON/plain/rich report distinguishes those receipts from current health and task completion. team show exposes both native journals, including partial channel creates when a later launch did not complete. Inventory, status, down and reset distinguish returned mixed deployments from unsettled attempts. Retained verification and guarded removal use the paths below.

Fresh boot previewCopy Anchor Link

team boot <file> --dry-run shows the resolved fresh topology, its target root, requested applications, roles, models and channels with declared delivery, replay and window settings. It shows the manifest's defaults: too, and a channel field that inherits one is marked (default), because boot seeds them as the space defaults. A setting neither declares is not inferred from a saved registry that this preview does not read. An open manifest with an explicit durable default or channel declaration is warned as unsupported: its native fresh apply refuses before effects because open mode starts no delivery daemon. Direct up -f --dry-run refuses the same contradiction. A space with no prepared or registered root previews the current folder. Its root source says whether that folder belongs to other spaces, and a hint names the cotal space create the apply needs: --root . for an unprepared folder, or a folder of its own when this one belongs to another space. It reads the local control directory, recorded spaces, maintenance journal, the selected manager PID record, possibly live process records, the root's additive deployment ledgers and attempts, and fresh boot attempts/summaries, plus selected-space service definitions, instances, supervisor receipts and retained native attempts. An unreadable additive or fresh record is listed rather than treated as absent. The team inventory separately lists unsettled fresh and additive attempts and returned agent/channel receipts. team show <id> distinguishes those historical records without inferring current card or broker state. Bare down --dry-run, team deploy --dry-run and both reset previews expose the relevant retained additive IDs and unreadable paths. The legacy bare down, down --preserve-state, fresh team boot and reset store refuse unsettled intents and unremoved agent/channel receipts before effects. Bare down and a later team deploy name cotal team remove <id> --dry-run, then --yes, for each deployment that blocks them. A complete matched static agent-retirement receipt releases only its original deployment's blocker. A bare up of a stopped controller stack starts it again over a returned channel-only deployment made on that stack, and keeps the deployment's records and cards as they are. Any other retained deployment, such as an unsettled attempt or an agent deployment, still refuses the start with LIFECYCLE_BLOCKED, and its Next names cotal team show <id> for each one. This is a known-blocker guard, not complete broker ownership. The service records belong to the workstation's selected-space store, not necessarily this boot root; their recorded state does not establish live ownership. A dead manager record remains visible, not proof that a later boot may replace it. For a concrete candidate broker address, the preview also lists other local registrations with that exact address. A match blocks fresh apply; no match does not prove exclusivity against aliases or remote tenants. A host-derived candidate leaves this comparison not checked. Where the manifest supplies a broker URL or uses the default without a host override, it makes a credless reachability probe. A failed probe is not proof of absence on a TLS-first listener. Its completeness is partial: port binding, exclusive local ownership, installed providers, credentials, grants and startup readiness have not been checked. It writes no record and starts nothing.

Fresh boot applyCopy Anchor Link

team boot <file> requires a prepared local root and starts its environment through the service controller. A root without a stack journal takes the fresh static-auth or open route. An eligible stopped stack can resume in static-auth, open or user-auth mode. The manifest's channels seed the broker. The stack's own manager starts its agents from a launch spec pinned by digest. One rich, JSON or plain receipt includes the controller members and requested topology. Agent joins are not checked, so the receipt is partial. A manifest that requests no agents leaves no join unchecked, so its receipt is ok. The fresh attempt and summary are launch records, not stop authority. No additive ownership ledger is created; cotal down uses the stack journal and each member's native lifecycle.

Before writes it checks the selected root, auth mode, broker address, runtime, connectors and retained ownership records. A journal, legacy PID record, possibly live component, service instance, supervisor receipt, native attempt, additive ledger or unsettled fresh attempt blocks a fresh boot. Unreadable records are never treated as absent. The lock protects local preparation, and ownership and bind checks run again before startup. A bind probe releases its socket and does not reserve the address. A later failure keeps its attempt and any started members, reports the known partial or unknown outcome and never repeats effects. A failed summary write also keeps the running stack.

team boot --foreground holds the stack's lifetime in the terminal. Ctrl-C or SIGTERM invokes the same guarded whole-stack stop as bare cotal down, without force killing members. Terminal loss or an external kill is not a confirmed stop. When a cotal down from another terminal completes, it prints Foreground stack released and exits 0, as a foreground up does, although its start receipt was partial. Foreground apply takes neither JSON nor plain output; a dry-run remains a finite preview. It does not establish native apply readiness or platform support. up -f --foreground still refuses on its retained launcher path.

Stopped-stack resumeCopy Anchor Link

A team boot in a root whose stack is up, with no cotal down since, refuses with LIFECYCLE_BLOCKED, says the space is already up and starts nothing. Its Next offers cotal team deploy <file>, which adds the manifest's channels and agents to the running stack, or cotal down before the boot. A recorded stack resumes only after its last whole-stack stop completed and its manager was never attempted. The manifest's broker.servers, broker.host and broker.auth must match that stack. A manifest that differs refuses CONFLICT (exit 2) and names each setting that differs with its recorded and declared values. Its Next names the manifest change that matches the stack, or a new space in another folder prepared with cotal space create <space>. The stack's broker inputs must be recorded, every applicable member must have its original settled stopped record, and no supervisor, retained custody or earlier team boot may remain. A stack that an earlier team boot started refuses with LIFECYCLE_BLOCKED, and its Next names cotal up in the root, which starts that stack again, and cotal team show <id> with each earlier boot's ID. The broker requires its kernel-exit receipt. Auth and delivery use their settled process stop and released custody; that is not proof of credential, lease or authority retirement. A generic process stop now records a process-exit receipt on the stopped instance when that stop signalled the pinned process and saw it exit. service stop --json and service status --json expose it as processStop, including the stop request, original process identity, signal and observation time. Human stop and status reports show the recorded signal, request and exit observation time, or Process stop evidence: not recorded. This is retained evidence, not a new signal from a repeated stop. An instance that was already gone gets processStop: null. Resume does not yet require this receipt. The resume receipt identifies the prior stop and each member's evidence. Changed records refuse under the lock. A journal without recorded broker inputs cannot take this route. Restart it once with its original existing --store-dir and its original --max-file-store and --host, if set, then run cotal down before the team boot. Omitting those values is safe only when the original start set neither. Preparation rechecks that explicit old store's directory identity instead of recreating it if it disappears. These checks do not exclude an unrelated local process changing the path. Open broker configuration stays inside the selected root's .cotal directory, even when its store is elsewhere.

For a first user-auth team, prepare with cotal up --user-auth --idp <url>, then cotal login, explicitly grant the invoking user with cotal auth grant cli --sub <subject> --full, and run cotal down before cotal team boot <file>. Do not start the manager between preparation and that stop. Before any write, that boot confirms your sign-in with its IdP. Without one, or when the IdP rejects it because it expired or was revoked, it refuses AUTH_REQUIRED (exit 2), says nothing was started or written, and Next names cotal login --idp <base>, then the team boot again. Any other failure of that check, such as an IdP that does not answer, is left to the boot. A static-auth or open preparation uses cotal up --no-manager before cotal down. No implicit owner grant or borrowed caller is substituted. A previously attempted manager is never adopted.

--space must agree with the manifest. Hand-registered and discovered spaces are not local boot targets. Public broker addresses and fresh-attempt views omit URL user info, queries and fragments; the private intent keeps the target. Native warnings and failures remain on stderr. Installed extensions may produce their own output outside the finite result.

Additive deployment verificationCopy Anchor Link

team deploy <file> --dry-run resolves the manifest and personas and requires a registration for the declared space in this root. It shows the declared channel delivery/replay/window values separately from its non-atomic live registry classification. If the selected running mesh is open, an explicit durable channel declaration is warned as unsupported by its missing delivery daemon; the read-only preview remains available. Additive deploy does not seed manifest-wide defaults. For an existing unmanaged card, it names only explicitly declared fields that differ from the live value, never the live instructions or description body. No declaration authorizes overwriting a saved card. It makes scoped, TLS-bound native reads of the presence and channel-registry buckets. A complete presence scan can prove an empty roster; a failed or malformed scan refuses rather than classifying absence. It classifies channel cards against the registry and agents against a unique prior ledger for the same manifest content. The derived membership feed, when readable, shows a lower bound of other readers, not authorization. These observations are non-atomic and do not reserve names or prove that a manager, runtime, connector, channel writer or launch grant will be ready. An unregistered or only prepared space refuses. --runtime changes the requested runtime; --allow-stale is apply-only and refuses in preview. Apply accepts all-new live channel-only manifests, or bounded all-new agent and mixed manifests on static/open meshes through one native manager. An identical settled channel-only deployment returns a no-op only after its recorded target and every original create revision match leader-served native reads. It reuses that full deployment ID without issuing a channel writer, creating an intent or changing cards. Missing or changed cards, ambiguous or unsettled attempts, and removed ownership cannot become this no-op. Nothing is repaired or adopted. The result names its observation time; it does not establish current readers or health. For an identical settled agent or mixed deployment, reapply verifies the original manager and every original incarnation. Static-auth uses active live slots; open meshes use the pinned manager's exact identity and lifecycle rows. Mixed cards must retain their original create revisions and content. Missing, retired or replaced agents refuse; nothing is relaunched or adopted. Persona drift refuses unless --allow-stale <names> names the stale agents. Those agents keep their old policy and receive a warning; the waiver never restarts or updates them. The no-op records an observation time and labels its ordered reads non-atomic. Unsettled, ambiguous or removed ownership cannot become that no-op. User-auth agent launch and retained verification remain unsupported. Runtime requests must match the original incarnations; no runtime or manager replacement is inferred.

Agent retirementCopy Anchor Link

team remove <full-deployment-id> [--dry-run] [--yes] can retire a returned static-auth agent-only deployment with at most 100 original incarnations. It requires the original registered root, broker, TLS mode and exact manager. The revision-32 native slots query checks each recorded name, principal and lifecycle UID. A successor, omitted slot or unconfirmed transition refuses before a stop. A dry-run only reads this impact; apply requires a default-cancel terminal confirmation or --yes in scripts. Every team remove --dry-run exits 0 with completeness partial: it reads the manager's slots and the card revisions in order, not atomically, and the membership feed is display-only, so the preview does not establish what the apply will find. The apply reads the slots and revisions again before it acts.

Under the root maintenance lock, apply records one private synced removal intent bound to the original creation receipt. Each necessary stop has a synced original request marker and an expected principal/UID guard. An accepted stop is not completion. The CLI observes that same incarnation under a 30-second loop budget, in addition to each native read's request bound, checking its native retired slot, retired lifecycle head and cleanup-complete fact. It retains both revisions and the answering manager epoch before continuing. An already retired incarnation needs no repeated stop. A missing row, successor, interruption or observation failure stops the sequence without retrying effects.

Only a complete matching retirement sequence yields the create-only full removal receipt. Earlier confirmed retirements and stop requests remain in rich/plain/JSON output on a partial or unknown result. An unsettled removal intent stays blocking and is never resumed by repeating its effects. team show keeps its original creation and removal journals, and inventory/status/down/reset distinguish historical retired receipts. It lists a card removal's steps in the order they were written: the cards in manifest order, and each card's delete submission before its outcome. The full receipt releases only this deployment's known additive blocker, not any other service, broker, shared-root or process guard. It never authorizes stopping current alias holders. No channel, broker, service or remote account is removed.

Channel-card removalCopy Anchor Link

team remove <full-deployment-id> also removes the registry cards of a returned channel-only deployment, with --dry-run for a read-only preview and --yes for scripted apply. It pins the original receipt, root, broker, auth mode and TLS requirement, and refuses changed card revisions, conflicting retained ownership or unreadable records. A changed card refuses CONFLICT naming each changed card. With no registration for the space, as after the stack's cotal down, the refusal names cotal up in the deployment's root, which registers the stack again, and then the same team remove. The preview discloses delivery and replay fallback to space defaults and removal of description/instructions. A derived membership feed is display-only and never authorizes or blocks this metadata change.

Apply records a private intent before any delete. Each card delete is conditioned on its original create revision, then read back as a later removal marker. Already absent cards name the remover unknown. Missing acknowledgements remain unknown with a blocking intent and no retry. A complete receipt releases only that deployment's retained blocker. History, grants and readers are unchanged; the channel stays addressable and no reader exclusivity or complete cleanup is claimed. Repeating a completed removal returns its historical receipt without deleting a later same-name card. A new card intent records the registry stream's name and creation time.

team remove <full-deployment-id> also removes what an unsettled deployment attempt created, such as one whose agent exited on launch, when the attempt's own journal records every effect. In manifest order, each channel and agent reached its accepted step or was never submitted, and no member has a step after one that was not submitted. An accepted agent counts by its native allocation, whether or not it joined. Removal then takes the route of a returned deployment of the same shape, with the same preview, confirmation and receipts. Each allocated incarnation is retired through the original manager by its identity and lifecycle UID, and one the manager already retired and cleaned needs no stop. Each card is then deleted at its accepted create revision. Agent retirement still needs static auth. The removal is bound to a digest of that journal, which is derived on each read and never written, so a journal that changes later no longer matches it. A channel or agent submitted with no recorded acceptance, an unreadable record of the attempt or a journal with no step leaves the effects unknown. team remove then refuses with LIFECYCLE_BLOCKED and names that step.

Unsettled removal reconciliationCopy Anchor Link

team remove <full-deployment-id> on an unsettled removal intent reconciles it from native observation instead of refusing. --dry-run shows each original member's classification, and --yes records only what the observation proves. It never sends a stop or delete, never repeats an unknown action and never continues a member that was not submitted; that member is reported unattempted and keeps the deployment blocking.

• A card whose delete was submitted without a confirmed outcome is read with the deployer view from the registry stream the intent names, between two stream identity reads. A later removal marker, or no message at all, records observed-removed with the remover unknown, and it counts toward the receipt. A later value records observed-superseded; that card is not owned and the deployment stays blocking. A card still at its create revision stays unknown, because the original conditioned delete could still be applied. An older revision, a revision past the stream's last sequence or a recreated stream refuses. A read failure is never treated as absence. An intent written before stream binding keeps its submitted cards unknown.
• An incarnation whose stop was submitted without a confirmed outcome is read with the privileged control instrument, which holds no stop or attach rows, from the original manager's exact-name slots. Only the original lifecycle UID's retired slot and head with cleanup complete record a retirement, under the original stop request. A terminal in progress stays open. A live original, an active slot with no live process, a retired slot without cleanup proof and an alias that now names a successor stay unknown. No process or presence absence settles anything. Open and user-auth deployments are unsupported.
• A mixed deployment reconciles retirement first. Its card intent is observed only after a complete retirement receipt, and a retirement settled by reconciliation does not start a card phase.

A receipt is written only when every original member settles as removed or retired. The receipt then releases that deployment's blocker as an ordinary removal receipt does.

Legacy removal previewCopy Anchor Link

team remove <full-deployment-id> --dry-run can also select a validated legacy ledger. It refuses a fresh boot ID or a changed root, broker, auth mode or TLS requirement. The old ledger removal path cannot adopt a new receipt. For a legacy ledger, a scoped deployer read asks the serving manager for a bounded native PS snapshot and compares recorded IDs, names and lifecycle UIDs. It reads current KV revisions for ledger-named channels without exposing their descriptions or instructions; an unavailable registry read refuses rather than treating cards as absent. The old ledger has no creation revisions, so a current revision and a matching name do not authorize deletion. It also reads the derived membership feed when available and shows other observed members on owned channels; that feed is display-only and cannot authorize channel removal. A missing or unusable manager PS refuses rather than calling owned agents absent. On a user-auth mesh a deployer view the identity provider refuses is PERMISSION_DENIED (exit 2), with the provider's sentence as the remedy. An unavailable membership feed stays explicitly unknown. Other retained or unreadable additive attempts under the root are shown separately; their effects are not attributed to this ledger. These native reads are not atomic. The preview changes nothing and cannot authorize a stop. Legacy removal still refuses effects, including with --yes. Agent retirement remains static-auth only. Channel-only metadata removal uses the existing scoped writer on static or user-auth meshes, or the open mesh's bare connection; it grants no new access.

Mixed deployment removalCopy Anchor Link

A returned static mixed deployment uses one preview, confirmation and maintenance lock. It first retires the original agent incarnations through their native manager. Only a complete retirement receipt permits fresh card checks and deletes conditioned on the original creation revisions. Both receipts are required to release the deployment's blocker. A partial or unknown phase stops there without replaying its effects. An interrupted transition can continue only after complete retirement and before any card-removal intent exists. Open and user-auth mixed removal refuse.

Manifest file checksCopy Anchor Link

team check <file> resolves the native manifest and referenced personas without a broker. It lists the manifest's defaults: and each channel's delivery class, replay and replay window. A field the channel leaves out shows the manifest default it inherits at boot, marked (default). A field neither declares does not claim a live registry value; the native seed may merge over saved metadata. The JSON result carries defaults beside the channels. --graph shows effective channel access edges from merged agent policies, marking persona-inherited channels that have no manifest card. It also shows each agent's role, capabilities and whether capabilities came from the manifest, a persona or neither. Terminal tables show up to 100 rows of each kind. The JSON result includes all resolved sources, merged policies, inherited scopes and, when requested, the channel graph, without persona bodies, kickoff prompts or opaque launch-option values. A manifest it cannot resolve, such as one whose space: is not a name space create takes, is refused with SCHEMA_INVALID (exit 1). The refusal names each problem by file, line and column where the parser has them, the key path, and the reason, and --json lists every problem under target.issues. A missing required key, an empty file, a value of the wrong shape and one outside a key's fixed values each get a plain reason. An unknown key that misspells a required key left out beside it, such as chanels, is one problem that names the key meant. A problem in a referenced persona names that file and the parser's reason without quoting its contents. A persona file that does not exist is named at its entry's line along with the manifest's other problems, so fixing those does not uncover it later. A manifest file that does not exist is refused with TARGET_NOT_FOUND (exit 2), naming its path. A folder given as the manifest, such as team check ., is refused with USAGE_INVALID (exit 1), naming the path and that it is a directory, as topology view refuses it, and --json gives the path as target.file. team boot and team deploy refuse a manifest they cannot resolve the same way, before they contact a broker. This file check does not prove live authorization, installed applications or launch readiness. These read views do not apply a boot, an additive deployment or a removal.