A stop that may already have signalled the host (a timeout or an
unexpected error) is trailed as failed, not refused; only exit 3 comes
before the signal. The response's stopped flag follows what busStop
reports, so a host that exited on its own between the read and the stop
is not reported as stopped by this press.
POST /api/orgs/ID/stop/bus runs busStop the way mosaic bus stop does: a
graceful SIGTERM close that leaves no host.lock or writer.lock, so an
offline schema migration can open the store straight after. It follows
restartBus for authority, same-origin checks, trail and receipt. A bus
already stopped answers done with state stopped and changes nothing; a
host identity the Console cannot prove refuses with every file unchanged.
The organization header gains a Stop bus button beside Restart bus, one
press, sharing its receipt and busy state. The restart icon now lives
inside its button, so a wide receipt no longer pulls it off the button.
Security review P2s:
- A binding proposal keeps only its structured request fields. The
free-text reason is never written to the private proposal or covered
by its digest.
- A root copy's receipts are checked at prepare, at check, and again
under the organization lock before the write and before publication.
A copy removed, replaced or unreadable since boot refuses with
instance-fired, copy-stale or root-policy-unreadable.
Build review P2:
- A root copy's admitted projects come from the project unions,
recomputed from the current document and the boot's root agents. A
project that excludes the copy, or whose union refuses, is never in
reach; a request naming it refuses with credential-scope-exceeded.
Build review P3s: tests for proposal-time shadowing, the loader's
copyBindings validation, a pending record that covers every service,
a stricter path check on the refusal table, and refusal codes on every
bare rejects in the binding tests.
CB0: registry incarnation minted at reserve (legacy entries get one at
the first binding proposal, under the registry lock); copyId, source
and definition pin, binding generation; a pure resolver where
effective = role intersect approved and only a bound, matching,
unshadowed record resolves.
CB1: copyBindings in the organization file (ids, scope names, project
ids and integers only). applyCopyBindingDecision is its only writer,
behind a module-private capability every other commit is fenced
against. A binding is an owner-only subtype of the hire decision:
only the chief raises it, no autonomy level, resolver agent, member
right, tester or reader resolves it. Digest, expiry, single
consumption, revision, generation, incarnation and definition pin are
rechecked before any effect; a bind writes pending, an unbind the
unbound tombstone, receipts are paths only. Nothing writes bound, so
nothing launches on a binding yet (CB2/CB3). The org service and the
Console update path keep copyBindings server-held. Eight new refusal
codes with next steps.
The agent-run refusal on a dev-local-owner start now checks the starting
process's own environment as well as the environment given for the
Console, so a caller cannot pass a clean env to get past it.
The start lock now names its holder (pid and start time). A detached
start that times out and kills its supervisor's process group clears the
lock only when it names that supervisor and the supervisor is gone, so a
forced kill no longer blocks every later start.
Build review P2: the supervisor ignores a lost reply pipe and keeps
supervising. Security P2: a stop signal during start is held until the
start settles, and the CLI's start timeout signals the supervisor's
process group. Security P1 (part): console start --dev-local-owner
refuses inside an agent run. Tests cover review, confirm and reverse on
the org autonomy route; TOOLS.md and the webui README are corrected.
A --dev-local-owner Console's organization pages read writable:false because
the per-org options had no autonomy writer and the route fell back to the
request transport, which org options never set. The org options now carry an
autonomy writer on the organization's console transport, fenced to the four
owner autonomy verbs. The route prefers it and keeps requestWrite otherwise.
mosaic console start --dev-local-owner --detach now runs the Console under a
detached supervisor, so Restart Console (exit 75) relaunches it on the same
address. A non-dev detached Console is unchanged. A deployed --sso --send
Console has no organization routes and keeps its root autonomy transport;
a test pins both.
- busRecover: a live pid with the recorded start time is the recorded process, so it refuses
("bus host identity is uncertain") whatever its argv. It used to read a mismatching argv as a dead
host. The pinned recovery test now covers a reused pid (live, other start time) as the wrong
process, and a new test covers the refusal.
- busRecover clears through setAside: each inspected entry is renamed aside and checked by device
and inode; a swapped entry puts everything back and refuses before any receipt or deletion, and
the deletion names only the aside entries.
- mosaic bus recover refuses inside an agent run (refuseInsideAgent), like mosaic org reload. The
Console-start revive does nothing when the Console's own environment carries an agent marker.
- Tests: org-recovery (identity, agent refusal), org-revive (setAside swap), bus-revive (markers);
the other reviveBuses tests pass an explicit env.
- The Console-start revive asks busRevive for pm:false; reviveStartOptions holds the rule (never
wider than the dead host's record) and is tested on its own. A dropped --pm is logged with the
way back.
- A Restart bus press waits for that organization's start-up revive (revivePending) before it reads
the bus, so it no longer races it for the lifecycle lock; a press that still meets org-locked gets
409 bus-busy, nothing changed.
- busRevive marks an error from busStart after a successful clear (error.recovered); the trail, the
Console log and the press response then say "failed", not "refused". A non-CliError logs its code.
- Tests: bus-revive.test.mjs (new), org-admin and org-revive cases for the above.
A Console unit stop (KillMode=mixed) SIGKILLs the bus hosts the Console
started, leaving host.json, host.lock and the socket behind, and the
dev Console's Restart bus refused 409 on that stale state.
- busRevive (org-bus): busRecover, then busStart detached on the
options the dead host recorded. busRecover still refuses a live host
pid, a lock that is not the dead host's, a live writer pid, or a
socket that answers or is uncertain.
- reviveBuses (webui): when a dev-local-owner Console starts, a bus that
answers is left (reattached); a bus with stale state goes to busRevive
with pm:false, trailed as bus.revive before anything changes.
- Restart bus (org-admin): for the exact dev owner only, a stale-state
refusal goes to busRevive; a signed-in owner keeps the refusal.
- The receipt reads "Dead bus host cleared; bus started".
A replaced installed harness answers null for countsAccount until the
bus restarts and reads the new module. A runtime test pins the
startBroker wiring for claude-code (covered) and pi (refused). The
plugin README says countsAccount also decides authority coverage and
is read once at start.
authorityBudgetCovered named claude-code to decide whether an account
quota covers every future launch, which fails the row 65 scan. The bus
host now passes each harness module's own countsAccount answer, read
once with its levels in harnessLevels, through autonomyHost to the
broker. Without the lookup no account quota counts (fail closed); the
provider/model branch is unchanged.
The catalogue refusals no longer tell every user to press Disable model
proxy catalogue. The server text names the proxy setup step. The Routing
page adds the Disable step only when cliproxy is listed and every moved
model comes from another listed provider, and sends a tester to the
owner. The switch mismatch message no longer claims a setting changed.
Tests: a fallback thinking-only change saves with no catalogue; a move
between model-proxy models refuses before and after cliproxy is removed.
A fixture-only organization reads a ready catalogue with no model proxy
client: the catalogue asks the proxy only when a module calls it.
A thinking-only change, and a limits patch that keeps harness, model and
entry, no longer asks the catalogue. A change that truly needs it still
refuses, and the WebUI names the next step: keep the model and change
thinking, or press Disable model proxy catalogue. The Routing switch
follows whether cliproxy is listed and removes or adds only cliproxy.
D4 review fix round.
P2: the org inbox took the hire form's HTML before its workspace read, so typing or a
submit result that landed during a slow read was painted over with the older draft.
inboxPage now takes a function and calls it while building the page after the read.
Two browser cases hold the inbox read: typing during it, and a refused start that
resolves during it, both survive the repaint (both fail on the previous commit).
P3: a hire start passes a bus code through only when the broker raises it before it
records the decision (hire-decisions prepare: the model route codes,
definition-not-hireable, definition-shared, no-chief-of-staff). Every other code,
including revision-conflict-revoked, request-timeout, invalid-envelope,
incomplete-request and codes the resolve or apply steps can raise, stays
outcome-unknown. Messages are worded for hire, fire and change alike.
D4, readiness P0-1 and P1-2.
On an org Console, #/inbox now carries the hire, fire and change form
below "Answered today". It posts to the current organization's bus
through the same decision endpoint and verb (hire.start.console); the
organization field is read-only. Inputs carry focus keys, so a poll
repaint keeps the draft, focus and caret.
A hire start the bus definitely refused now keeps its own code. The
model route codes (model-provider-unavailable and its kin) get a next
step. Any other definite code is named, and the message never claims
that nothing was recorded. A lost or unreadable reply is still
outcome-unknown. The templates, confirmation and resolve steps are
unchanged.
Re-check 3 found connector.start() still quoted the journal directory in
its STOP refusal, and the CLI passes a DiscordError through unchanged.
It now uses journalLabel like the CLI and journal refusals. The connector
test asserts the path is absent: 22/1 on the old source, 23/0 after.
The discord CLI's top-level handler prints a native Node error that is not
already a labelled refusal as 'cannot <syscall> <label> (<CODE>)'. The journal
and approvals readers and writers name the file by label at the source, so
conversation, control-board and webui callers get the label too. The session
directory check moves to engine-pi.mjs as sessionContinues(). Exit codes and
refusal codes are unchanged.
A token file, context file, tool root, data root or setspark key file
that changes between its check and its read failed with a native error
whose message carried the host path. Each read now names the file by
label and code (quietRead): same error object, type and code, so exit
codes, refusal codes and check order are unchanged.
The journal label import pulled binding.mjs and its config readers into
every journal.mjs importer. The queue test repo copies only journal.mjs
and errors.mjs, so its genesis failed with ERR_MODULE_NOT_FOUND (webui
reads.test.mjs red, 4 tests). BINDING_NAME now lives in errors.mjs and
binding.mjs re-exports it, so its export is unchanged.
The board API returns a scan refusal in its JSON error body. The test
starts a board on port 0 over a seen.json that is not an object and
checks the body names seen.json by label, with no sentinel or scratch
path. It fails on BASE, where the body quotes the host path.
Review "changes 8" fix round for org-redact6.
- launcher: a harness module's throw names its system code or class,
never its message; preflight, loader, bundle, checkout, sandbox and
route text is relayed with host paths replaced (relay()).
- claude-code and pi preflights name MOSAIC_CLAUDE_BIN and the
repository's node_modules, never a path; persona and skill bundle
errors and sandboxProblem name their file or the sandbox by label.
- seat session: thread store, legacy store, overlap, busy lock and
newRun refusals name the place under the data root.
- discord: unknown keys, binding names and turn ids are described when
not plain; approvals, journal and repository read errors name the file
and code; git drops stderr lines that carry a host path and a spawn
failure names its code only.
- seat, control-board: registration fields and seen keys are described
when not plain; invalid seat name and layout state the constraint.
- organization: the registry names the config directory and registry
file by label; --project is quoted only when it is a plain name;
systemConfigProblem moves to its own module.
- scripts/mosaic-config.mjs: the config by where it came from, the data
root and its marker by role, keys and arguments only when plain, and
error codes in place of raw messages. Exit codes unchanged.
- webui sso-config: an unknown console key is named only when plain.
Sentinel tests in each package's path-redaction-labels file fail on
BASE; pins that quoted a path now assert the label and its absence.
redact6. A refusal names the field, the file or the instance. It never quotes a
config value, a host path, or a raw e.message or stderr that carries one.
- organization: checkCredentialRef names the field and each forbidden root by
label; OrgError keeps .path only for a document field ($...); the registry,
fence and lock refusals use the quiet helpers (call and code).
- cli: loadSystem never relays mosaic-config stderr: instance refusals keep
their code and names, everything else reads as a field and its constraint
against a labelled config. Project-key, access warning, halt gate, sessions
file, launch socket, harness relays and the workspace rename name labels.
Launch-log reasons read a Node system error as its call and code.
- scripts/mosaic-config: configuration not found names where the config came from.
- discord, seat, control-board: config, token, context, tool-root, journal,
binding, registration and seen-marks refusals name labels.
Exit codes, refusal codes and check order are unchanged. Each new or changed
test fails on BASE be698e602.
A4's hire writes an org:hire-<uuid> definition, so the chief's in-session
launch of its hire now reaches the launcher's held-secret check. The fixture's
op table lacked definitionScan and refused invalid-host-request; process.mjs
answers it.
The hire service takes the bus's definition check as scanDefinition and runs
it on the stored proposal entry before preview, check or apply goes further,
and on the approved entry again just before the writers. An approval whose
definition holds a credential the store came to hold after the raise refuses
definition-holds-credential before the decision is consumed; nothing is
written. Without the hook the entry still goes through assertClean whole.
Bus start and credentials.issue also check the boot copy snapshot's
definitions (A4 root agents file definitions, re-applied on every reload),
which no organization file holds.
credentials.issue runs a check after it normalizes the value and before the store, the
organization file or memory changes: the value must not appear in the definitions of any
organization the bus serves (instructions, personality and every other field or key). A hit
refuses definition-holds-credential naming the organization, definition and key, never the
value or the credential. An organization file the check cannot read refuses
definition-scan-unreadable (new TI entry). The check runs on the bus loop inside issue, so a
launch admission can no longer pass and then see its definition text become held before spawn.
The final admission re-scan stays as the backstop for values held outside the verb.
Review round (P2): a credential issued between the launcher's definition scan and
spawn could make text already in the definition a held value; a same-date rotate
leaves the organization file's revision unchanged, so the revision fence missed it.
The bus's final request.admission for an instance, the last bus op before spawn,
now runs the held-secret check on the org: definition the instance names, on the
bus loop that issue runs on, and refuses definition-holds-credential.
Self-found: check() takes the scanned entry from the same read as resolveInstance,
before any await can let another check replace the shared organization.
T1 follow-up. A definition's text becomes an agent's contract, but a load ran the
held-secret check with an empty store. Bus start now checks every definitions entry
(the boot copy's org: entries and the file's own) with the live store, and the
launch check asks the bus the same question for the org: definition it just read,
before a contract exists. A hit refuses as definition-holds-credential, a new code
in the refusal table; the refusal names the definition and key, never the text or
the credential. bus start records the refusal in the boot status.
Review P1: the CLI wrote definitions with a no-op secret check. setDefinition
now refuses definition-secret-check-required when a write that adds or changes
an entry has no assertClean. New bus verbs definition.import|invent|edit|remove
run the writes for the business's human at the cli with the credential
store's assertClean; the tester, agents and readers refuse. The CLI's
definition writes call the bus and map its codes to messages that never
quote the definition.
Roles round 2, slice T1 (roles-and-rails design sections 4, 5 and 13).
- The org file's definitions key: each entry {rev, from?, role, instructions,
personality?} passes checkComposition; a bad entry refuses the whole file,
naming the definition and key, never a value.
- org:<name> instance references resolve to the file's definition; a missing
one refuses definition-missing, naming the instance. Template names are
unchanged.
- Template identity is the name plus a sha256 over the template JSON and its
contract; templateHashes reads the current hash for each template.
- The writer's setDefinition is the only path that changes definitions: it
checks the composition, bumps rev, refuses agent actors, and refuses
definition-in-use for a definition an instance names (A4 adds the change
decision). write() refuses organization-definitions-protected.
- Definition verbs import, edit, invent and remove need org.hire
(definition.*.console in VERB_RIGHTS) and refuse every agent caller.
- The diff read lists the primitives that changed and a line diff of the
instructions. It applies nothing.
- CLI: mosaic org definitions templates|list|show|diff|import|invent|edit|remove.
The review round changed the config-command relay and the org document
refusal to name the file by label. cli/tests/json-redaction.test.mjs still
required the host path in those two CLI refusals. They now require the
label and the path's absence. The direct scripts/mosaic-config.mjs checks
and the loadSystem check (PL1, deferred) still require the path.
Sol review (org-redact5, changes 6):
- org-service: Node filesystem errors become "cannot <syscall> <label> (<CODE>)";
code, syscall, deleteStep and deleteReceipt stay. git failures name the verb
and exit code, never git's stderr.
- org-console: consoleStop names "another checkout", not the entry path. Exit 3.
- config-command: the system config is named by where it came from. Exit 3.
- cli org: documents and raw fs errors are named by kind and code. Exit 2.
- instances: --vars file named by flag and kind. Exit 4.
- hosts: a refused command is named by the basename refusedCommand matched.
- role.mjs (T1): the change is back to the readJsonFile label only.
Sentinel tests for each; pins in last-values, rename-b-commands, org-recovery.
PR #1916 rule: a refusal names the field or file, never a config value
or a host path. readJsonFile's three refusals and parseJsonFileText now
name the file by its `what` label only. Callers whose label alone is
ambiguous add a name that already passed a strict pattern (business id,
role name, org id, turn id, project key) or a fixed path under the data
root. Exit codes, refusal codes and check order are unchanged.
organization: util, access, project, config-command, library-seats,
org-service deleteOrg receipt, agent-levels regex; T1 files role.mjs and
organization-config.mjs change only the readJsonFile label.
cli: access lock, accounts owner, host state/lock, notifier journal,
notify-config, run-evidence, dev-fixture, git-socket, org-bus, org-console.
discord: turn record label names the turn id.
Sentinel tests (sk-test-NOT-A-SECRET-1234 as a path component) cover each
changed refusal and fail on BASE; old path pins updated.
The row299 root-side pair (mos-1409) spans supervisor start-up and
clean_session, which scans every /proc entry on the host. With about
1,850 processes that adds 0.6 to 1.6 s, so the unchanged inner probe read
2.5 to 3.9 s and failed the [1.25, 2.75) band, while the 1 s inner-timer
mutant read 2.06 to 2.38 s and passed. Widening the bound alone would
have kept that mutant alive.
A second observer pair, also confined to the private test copy, brackets
only the inner timer: begin just before the supervisor starts its sleep,
end at the start of finish(). It writes only for the SO_PEERCRED peer and
at most one pair. The unchanged band now applies to that interval
(2.01 s on base, 1.00 s and 3.00 s for the 1 s and 3 s mutants, so both
fail). The root pair remains required and must contain the timer pair;
probe-inner-timeout already places it before the outer bound.
jason.woltje