@awebai/oats 0.69.23 → 0.70.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (88) hide show
  1. package/bin/oats.mjs +112 -20
  2. package/docs/capabilities.md +697 -11
  3. package/docs/capability-manifest.schema.json +66 -0
  4. package/docs/configuration.md +27 -7
  5. package/docs/design/2026-11-19-sandboxes.md +595 -121
  6. package/docs/design/2026-11-20-provider-home-lifecycle.md +322 -0
  7. package/docs/design/README.md +7 -3
  8. package/docs/desktop-cli-api.md +330 -9
  9. package/docs/desktop.md +181 -48
  10. package/docs/execution-targets.md +210 -4
  11. package/docs/first-team.md +11 -0
  12. package/docs/implementation.md +24 -3
  13. package/docs/integrations.md +2 -2
  14. package/docs/knowledge-capability-authoring.md +1 -1
  15. package/docs/knowledge-reference/proposal.md +10 -0
  16. package/docs/knowledge.md +20 -11
  17. package/docs/official-catalog.md +2 -2
  18. package/docs/packages.md +7 -7
  19. package/docs/release-notes/v0.69.24.md +264 -0
  20. package/docs/release-notes/v0.70.0.md +155 -0
  21. package/docs/souls-and-instances.md +57 -0
  22. package/docs/trust-model.md +8 -1
  23. package/docs/tui.md +21 -1
  24. package/docs/upgrading.md +19 -5
  25. package/docs/workspaces.md +13 -5
  26. package/lib/automation-scope.mjs +1 -1
  27. package/lib/capability-show.mjs +20 -1
  28. package/lib/command-flags.mjs +2 -2
  29. package/lib/compile-cache.mjs +1 -1
  30. package/lib/core.mjs +1455 -203
  31. package/lib/errors.mjs +4 -1
  32. package/lib/home-integrity.mjs +329 -0
  33. package/lib/hook-env.mjs +107 -0
  34. package/lib/host-run.mjs +11 -3
  35. package/lib/instance-events.mjs +32 -5
  36. package/lib/instance-git.mjs +9 -0
  37. package/lib/instance-inspect.mjs +107 -9
  38. package/lib/instance-lifecycle.mjs +15 -3
  39. package/lib/instance-listing.mjs +101 -8
  40. package/lib/instance-resolution.mjs +85 -2
  41. package/lib/instance-upgrade.mjs +170 -14
  42. package/lib/kernel-features.mjs +10 -1
  43. package/lib/node-version.mjs +9 -0
  44. package/lib/provider-call-child.mjs +60 -0
  45. package/lib/provider-resume.mjs +83 -0
  46. package/lib/provider-test-seams.mjs +20 -0
  47. package/lib/read-call.mjs +2 -0
  48. package/lib/remote.mjs +33 -1
  49. package/lib/reserved-launch-env.mjs +8 -1
  50. package/lib/retire-admission.mjs +12 -1
  51. package/lib/sandbox-allowances.mjs +329 -0
  52. package/lib/schedule.mjs +14 -4
  53. package/lib/servers.mjs +38 -5
  54. package/lib/session-inspect.mjs +123 -87
  55. package/lib/session-providers.mjs +575 -0
  56. package/lib/session-receipt.mjs +93 -0
  57. package/lib/session-resume.mjs +68 -15
  58. package/lib/stable-node.mjs +3 -0
  59. package/lib/triggers.mjs +3 -1
  60. package/lib/upgrade-state.mjs +6 -1
  61. package/lib/worktree-hooks.mjs +18 -2
  62. package/lib/worktree.mjs +21 -7
  63. package/package-catalog.json +2 -2
  64. package/package.json +2 -1
  65. package/packages/client/capability-list.mjs +21 -0
  66. package/packages/client/cli-adapter.mjs +19 -4
  67. package/packages/client/cli-locator.mjs +5 -1
  68. package/packages/client/deployment-data.mjs +23 -1
  69. package/packages/client/deployment-panel.mjs +25 -3
  70. package/packages/client/home-board.mjs +33 -19
  71. package/packages/client/inspect-boundary.mjs +18 -8
  72. package/packages/client/instance-acts.mjs +30 -3
  73. package/packages/client/instance-board.mjs +206 -0
  74. package/packages/client/instance-presentation.mjs +55 -0
  75. package/packages/client/liveness.mjs +15 -4
  76. package/packages/client/own-codes.mjs +5 -1
  77. package/packages/client/remote-roster.mjs +6 -1
  78. package/packages/client/session-backends.mjs +31 -0
  79. package/packages/client/session-start.mjs +18 -6
  80. package/packages/client/spawn-apply-cli.mjs +5 -3
  81. package/packages/client/spawn-apply-contract.mjs +25 -2
  82. package/packages/client/spawn-apply.mjs +5 -1
  83. package/packages/client/spawn-messages.mjs +3 -0
  84. package/packages/client/terminal-contract.mjs +4 -0
  85. package/packages/client/waiting-on-you.mjs +12 -2
  86. package/packages/tui/lib/model.mjs +14 -3
  87. package/packages/tui/lib/store.mjs +12 -3
  88. package/skills/oats-getting-started/SKILL.md +3 -3
@@ -80,7 +80,7 @@ and the rules its tests pin.
80
80
 
81
81
  | path | contents |
82
82
  |---|---|
83
- | `bin/oats.mjs` | the CLI: argument parsing, JSON envelopes, the verbs. Its only static imports are Node built-ins and two dependency-free modules, `lib/compile-cache.mjs` and `lib/hot-reads.mjs`: it turns Node's compile cache on first, then imports the kernel and `yaml` with top-level `await import()`, in their order, because static imports are all parsed and linked before a module's body runs. A hot read (`version`, `session inspect`, `status`, `workspace status`, `instance events`, `instance waiting clear`, given only their own flags) imports only the leaves it uses, never `core.mjs` ([`hot-reads.mjs`](#kernel-modules)); every other command imports every module. A hot read takes its names from `lib/hot-reads.mjs`: a name newly used on a hot read's path must be listed there, under its import line, or its first use throws (a tripwire), and `test/cli-hot-reads.test.mjs` pins what each hot read must not load |
83
+ | `bin/oats.mjs` | the CLI: argument parsing, JSON envelopes, the verbs. Its static imports are Node built-ins and dependency-free startup helpers (`node-version`, `compile-cache`, `hot-reads`, `install-stamp`): its first statement checks the Node 22 floor, before the install stamp or compile cache; then it imports the kernel and `yaml` with top-level `await import()`, in their order, because static imports are all parsed and linked before a module's body runs. A hot read (`version`, `session inspect`, `status`, `workspace status`, `instance events`, `instance waiting clear`, given only their own flags) imports only the leaves it uses, never `core.mjs` ([`hot-reads.mjs`](#kernel-modules)); every other command imports every module. A hot read takes its names from `lib/hot-reads.mjs`: a name newly used on a hot read's path must be listed there, under its import line, or its first use throws (a tripwire), and `test/cli-hot-reads.test.mjs` pins what each hot read must not load |
84
84
  | `lib/` | the kernel (below) |
85
85
  | `injects/` | the kernel and work-mode instruction blocks composed into every instance |
86
86
  | `skills/` | bootstrap skills shipped with the kernel |
@@ -105,6 +105,7 @@ and the rules its tests pin.
105
105
  | `capability-show.mjs` | `oats capabilities show`: one catalog row's inject and skill files, read at its commit. Its text rules (`decodeText`, `skillDocDescription`, the limits) are also the rules of `inspect --soul --skills` (`instance-inspect.mjs` `soulSkills`), which reads a soul copy's own skills locally, and of `--composed-skills` (`composedSkillsIn`), which adds the capability skills of the scratch composition |
106
106
  | `materialize.mjs` | the shared home composer for spawn and module upgrade: `stageMaterialize` stages module copies and derived files, then a durable rename journal commits or rolls back a replacement in an existing home; `moduleSkills` is the one rule for which skills a module gives, read over the fetched tree, or over a remote listing for the spawn preview (`plannedModuleSkills`) |
107
107
  | `instance-upgrade.mjs` | `planUpgrade` and `applyUpgrade`: verified package-module staging, provider adoption, idempotent forward completion, undo and retirement baseline refresh for a stopped local home, using the shared composer for derived home files |
108
+ | `node-version.mjs` | the CLI startup check and `E_NODE_UNSUPPORTED` envelope for Node before 22, shared with unit tests. No imports or effects; loads on Node 18 and 20 before the install stamp, compile cache or kernel. CI runs the actual CLI on Node 20 (`scripts/check-unsupported-node.mjs`) to prove the refusal precedes every dispatch path |
108
109
  | `compile-cache.mjs` | the CLI's Node compile cache (awebai/oats#1630): `$OATS_HOME_DIR/compile-cache`, made 0700, unless the caller's `NODE_COMPILE_CACHE` or `NODE_DISABLE_COMPILE_CACHE` decides; any failure is no cache and prints nothing, and `NODE_COMPILE_CACHE` is never put in the environment children inherit ([configuration](configuration.md#the-compile-cache)). Node built-ins only; kernel-private, not in `./core` |
109
110
  | `hot-reads.mjs` | the CLI's hot reads (awebai/oats#1630), and the warm reader's start (`oats reader --stdio`, whose reads `reader.mjs` imports itself): each one's form (its words and the flags it takes) and, for each import line of `bin/oats.mjs`, the leaves it loads in that line's place; a line it does not list loads nothing, and any other argv loads everything. A name a read does not list is a tripwire: destructuring it is free, and any use throws a `HotReadUnlisted` defect naming the read, the line and the name. `test/cli-hot-reads.test.mjs` runs each read with a load hook and pins what it must not load. Kernel-private, not in `./core` |
110
111
  | `roster-read.mjs`, `workspace-read.mjs`, `read-call.mjs` | the deployment reads (awebai/oats#1630): `oats status` (`roster-read.mjs`: the roster, its drift and its workspace identity) and `oats workspace status` (`workspace-read.mjs`: members, packages, automations, the desktop facts, and the rows `oats sync` reports too), one implementation behind the CLI and a process that answers them many times. Each has an answer function (`statusAnswer`, `workspaceStatusAnswer`: `{ exit, answer, stderr }`, `answer` the JSON the CLI prints, a refusal included) and a data function (`statusView`, `workspaceStatusView`) bin's text renderer reads. `read-call.mjs` is one call's scope: it takes an explicit context (the absolute `dir`, `maxAge`, the caller's read session, the cache, an optional deployments observer), resets at its start every piece of module-level state the reads can reach (`READ_STATE`; `test/read-call-state.test.mjs` scans the reads' import closure and fails on a binding with no row), answers refusals as envelopes, and drains the session's notices. Within one call, two pieces of work are not repeated: `listInstances` hands the records it read to the rest of the read (`records`: `legacyCapturedHomes` and the waiting read take them instead of parsing them again), and `workspace status` asks git for each member clone's remotes once (`remotes`, shared by the automation scope and the desktop facts); both are maps the read makes and drops, never module state. A call registers no process listener, closes no session and reads no argv or cwd: the CLI passes `commandSession` (created at the first remote read, with its exit and signal listeners) and closes it at the end of the command, as for every verb. The read part of the dispatch is there too (`parseMaxAge`, `readBudget`, `callSession`); `test/read-call-equivalence.test.mjs` answers both reads many times in one process and compares each answer with a fresh CLI run's. Leaves: never `core.mjs`. Kernel-private, not in `./core` |
@@ -131,6 +132,10 @@ and the rules its tests pin.
131
132
  | `session-end.mjs`, `session-resume.mjs`, `session-resume-contract.mjs`, `session-seen.mjs`, `start-env.mjs` | session resume (#1559): what every local listing saw of each home and the pure rule that decides which losses the sweep resumes (`session-seen.mjs`: local listings and the sweep recording homes it started; never in the way of the listing); the contract's names (`sessionEnd`, the start's `resume` object, the sweep's outcomes and reasons, the feature) in one dependency-free module that every producer, test and doc reads (awebai/oats#1567 settles them); how a launched home's last session ended (`sessionEnd`, from the kernel-private exit marker and stop record, both forms of the marker), and the machine sweep `oats session resume` with its lock (a claim taken through `dir-lock.mjs` `acquireDirLock`, its record naming the boot too; `E_RESUME_LOCK_STUCK` for a lock that needs a person, which `oats doctor` reports as `resume-lock-stuck`) and its per-boot note (refusals, and the count behind the bound of 3 sweep starts per home per boot); a resume holds the home's lifecycle claim (`claim.mjs` below), as every start does, the sweep's status-row read of a retire or a claim holder is only its filter, and every stop OATS does writes the stop record through `core.mjs` `recordStop`, a retire's before its window kill; the record of each start's `fromEnv` and `fromFile` values (`start-env.mjs`: a keyed digest per source, never a value) and the one check a resume and `--plan` share, so a resume never starts under a changed value; the conversation a resume continues is `packages/record/lib/last-conversation.mjs` |
132
133
  | `stable-node.mjs` | the one rule for a node path written into anything that outlives the process: the host timer's unit and `OATS_NODE_BIN` ([capabilities](capabilities.md#the-node-a-hook-writes-down)). It imports nothing of the kernel |
133
134
  | `trigger-sources.mjs` | a capability's `triggerSources`: the use-time validator (per source, never from the manifest contract or discovery) and the source wire's answer and event rules; `triggers.mjs` runs a source with `instance-inspect.mjs`'s `runModuleCommand`, the provider-check runner |
135
+ | `session-providers.mjs` | a capability's `sessionProviders` (feature `session-providers`): the use-time validator (per provider, never from the manifest contract or discovery), `--backend`'s selection among an instance's capabilities, the wire (one call per verb of the kernel's own copy of the home's module, with the hook environment) and the rules of each verb's answer, the receipt's `runtime` for a provider home. A leaf: `session-inspect.mjs` and `instance-listing.mjs` call a provider on the hot reads; `core.mjs` holds the launch, stop, start and retire of a provider home. |
136
+ | `home-integrity.mjs` | what the kernel runs of a provider home ([capabilities](capabilities.md#what-the-kernel-trusts-in-a-provider-home)): its own copies of the home's modules, kept beside the receipt (`session-receipt.mjs`, the receipt's own primitives), the receipt's `composition`, and the capability rows read from it. A leaf: the hot reads load it. |
137
+ | `hook-env.mjs` | the environment a capability's code runs with in an instance home (`lifecycleHookEnv`, `teamEnv`, `stableSoulId`, `instanceSoulDir`, `kernelBin`): every lifecycle hook's and a session provider's commands'. A leaf, re-exported by `core.mjs`. |
138
+ | `provider-resume.mjs` | the resume cap of a home on a session provider: at most `RESUME_LIMIT` resumes of one home in `RESUME_WINDOW_MS`, in one boot, counted before the provider's input is called (the sweep's notice, `session start --resume`, any input to a lost session). One kernel-private file per home under `OATS_HOME_DIR`. A leaf. |
134
139
  | `refused-text.mjs` | the one set of characters the kernel never stores or prints in text it did not write (`REFUSED_TEXT`), and `safeText`, which replaces them |
135
140
  | `host-run.mjs`, `host-run-work.mjs`, `host-run-child.mjs`, `host-run-admission.mjs`, `host-stop-child.mjs` | `oats host run` (#1789), the foreground entry for a host with no timer (a container): its refusals before any effect (inside an instance session, an active host timer, another host run), its lock (a claim taken through `dir-lock.mjs` `acquireDirLock`, its record naming the boot, as the resume sweep's), the sweep at the start, then the tick and the sweep each minute, every one an owned child group (`host-run-work.mjs`) so that the loop stays free for the signal. On SIGTERM no minute is armed; active work gets 10 s, then the private scratch-file fence is closed and owned groups get 5 s of cancellation. The private bootstrap carries the fence only through kernel child argv, never public flags or environment. A detached kernel group atomically registers its pid/start and waits for the owner to verify it and acknowledge responsibility before importing launch-capable code; cancellation therefore does not depend on an intermediate supervisor remaining responsive. Schedule/trigger loops, resume batches and core launch boundaries read it synchronously, because a synchronous tick cannot service IPC while a command runs. Interrupted attempts remain unknown. The host-run claim stays held until owned work ends. After a final bounded listing, each running home is stopped by a private child of its own (`host-stop-child.mjs`: `core.mjs` `stopInstanceSession` with `withHost`, whose stop record asks for a resume and is written before the signal, read by `session-end.mjs` as `lost`). The clock, the timer, the children and the signal are options, for its tests |
136
141
  | `schedule-command.mjs`, `schedule-command-child.mjs` | synchronous scheduler adapter and asynchronous child supervisor: bounded output, TERM/KILL escalation and observed exit; no changes to synchronous tick/lock callbacks |
@@ -557,7 +562,12 @@ differs), `package-manifests` (a package's manifests), `list` (a skill
557
562
  listing) and `tree-oids` (the tree ids of a set of directories). An entry is only ever a pure function of those bytes and this kernel's
558
563
  code, never local state and never a transient error; it is written
559
564
  atomically, a corrupt one is a miss, and `pruneStores` bounds the store
560
- (`PARSED_LIMITS`, least recently used first). `--max-age` adds the observation
565
+ (`PARSED_LIMITS`, least recently used first). A read session made with
566
+ `offline: true` never reaches a remote: a head is the one this host last
567
+ observed (its record, of any age), a commit and a file are read only when the
568
+ cache holds them, and anything else is `E_REMOTE_UNREADABLE` with reason
569
+ `not-cached`. The sandbox allowances of a session provider's start are
570
+ resolved in one (`lib/instance-resolution.mjs` `homeSandboxSource`). `--max-age` adds the observation
561
571
  store `<cache>/.observed/` ([Observation reuse](desktop-cli-api.md#observation-reuse-feature-observe-max-age-oats-0311)):
562
572
  one record per (repo key, ref args, url digest), so two spellings of one repo
563
573
  keep a record each; the url itself is never written. Beside it,
@@ -967,7 +977,18 @@ inherits it (awebai/oats#816):
967
977
  names on the real one and gives the saved one back with
968
978
  `setProcessEnvironment` when it ends; everything else goes through
969
979
  `fx.inEnv`. `test/fixture-real-environment.test.mjs` refuses an assignment
970
- to `process.env` in any test script. The fixture's
980
+ to `process.env` in any test script. **A test writes an executable only
981
+ inside its fixture** (#2042): a harness stand-in, or a home's recorded
982
+ `launch.executable`, is written with
983
+ `writeFixtureExecutable(fx.base, path, content)`, which refuses a path
984
+ outside the fixture's base, one reached through a link, and a symbolic or
985
+ hard link at the path, with nothing written. `fx.inEnv` keeps the PATH of
986
+ the process that runs the tests, so an in-process spawn records the
987
+ harness it finds there, which on a machine with the harness installed is
988
+ the person's own binary: a spawn whose recorded executable a test will
989
+ rewrite runs with the fixture's PATH (`process.env.PATH = fx.env.PATH`,
990
+ put back afterwards; its inert `pi`, `claude` and `codex` come first), and
991
+ asserts `insideFixture(fx.base, launch.executable)`. The fixture's
971
992
  environment sets `HOME` and the XDG base directories under the base, and
972
993
  `GIT_CONFIG_NOSYSTEM=1`, `GIT_CONFIG_GLOBAL=/dev/null` and
973
994
  `GIT_TERMINAL_PROMPT=0`, so no host git
@@ -41,7 +41,7 @@ arrives from.
41
41
  # oats-workspace.yaml: one default per slot, for every soul
42
42
  packages:
43
43
  oats.okf: v6.3.0
44
- oats.aweb: v1.46.0
44
+ oats.aweb: v1.47.0
45
45
  oats.linear: v1.0.1
46
46
  oats.jira: v1.0.1
47
47
  defaults:
@@ -97,7 +97,7 @@ operations in full.
97
97
  - **`oats.aweb-tasks`** (in the `oats.aweb` package from 1.31.0) fills the
98
98
  tasks slot with aweb tasks, acting as the instance's own `oats.aweb`
99
99
  identity in its default team. **An existing workspace opts in** by pinning
100
- `oats.aweb: v1.46.0` (the feature needs 1.31.0 or later), adding
100
+ `oats.aweb: v1.47.0` (the feature needs 1.31.0 or later), adding
101
101
  `tasks: { oats.aweb-tasks: { from: package } }` to `defaults`, and running
102
102
  `oats sync`. Only new spawns get it. A soul that names its own tasks
103
103
  provider in `capabilities:` (`oats.jira`, `oats.linear`) then holds two and
@@ -29,7 +29,7 @@ an author soul:
29
29
  ```yaml
30
30
  # oats-workspace.yaml
31
31
  packages:
32
- oats.framework: v1.27.0 # provides oats.core, oats.setup, oats.knowledge-theory
32
+ oats.framework: v1.28.0 # provides oats.core, oats.setup, oats.knowledge-theory
33
33
 
34
34
  # souls/<author-soul>/soul.yaml
35
35
  capabilities:
@@ -51,6 +51,16 @@ so there is no queue, no background worker and no watermark to keep.
51
51
  record the checkpoint and its exact outcome in the instance's own record. A
52
52
  partial edit or an opened process is not a delivered proposal.
53
53
 
54
+ **Steps 2 and 5 may move to the reviewer.** A capability may keep the
55
+ proposer's consultation to a quick search and leave duplicates, the canonical
56
+ home and the destination node or subject to the reviewer. A proposal may then
57
+ name a node or subject the proposer does not own, flagged for the reviewer
58
+ rather than refused. `oats.okf` 6.3 works this way: its proposals carry such
59
+ findings as warnings the knowledge maintainer resolves before merging. Say in
60
+ the capability which side settles each of these, and make what the reviewer
61
+ must settle visible on the proposal. The exclusions and the screen before
62
+ delivery stay with the proposer: they protect what leaves the host.
63
+
54
64
  Never promote secrets, credentials, personal data about people outside the team
55
65
  (clients, correspondents, any other third party), third-party messages
56
66
  verbatim, tool noise, readily re-derived code descriptions or task-only plans.
package/docs/knowledge.md CHANGED
@@ -231,14 +231,15 @@ other than `off`, or a soul `knowledge.propose` key, refuses the spawn
231
231
  head of the same PR and each decision made while working are not
232
232
  checkpoints: those are captured in `notes/`, and the next checkpoint
233
233
  carries them. First update state, log and relevant notes. Apply
234
- `knowledge-theory` and search accepted knowledge (`oats okf search`, then
235
- `cat`): already accepted with nothing to add means no proposal. Make **one
236
- proposal per checkpoint, carrying only what no earlier proposal carried**,
237
- and record every checkpoint in `log.md`, either with its proposal's PR
238
- URLs or with `nothing to propose` and the reason. Claim only for owned
239
- nodes (or, for a version-2 soul, in `claims/` on its subjects); a lesson
240
- for a node the soul only reads is named in one line in its hand-back for
241
- that node's owner.
234
+ `knowledge-theory` and run a quick `oats okf search` for each claim: when
235
+ in doubt, propose, because duplicates, the canonical home and the subject
236
+ are the knowledge maintainer's to settle. Make **one proposal per
237
+ checkpoint, carrying only what no earlier proposal carried**, and record
238
+ every checkpoint in `log.md`, either with its proposal's PR URLs or with
239
+ `nothing to propose` and the reason. A claim goes in the owned nodes (or,
240
+ for a version-2 soul, in `claims/` on its subjects); a lesson that belongs
241
+ in a node the soul only reads may be proposed there too, and the check
242
+ flags it for the maintainer, who chooses its home.
242
243
 
243
244
  The instance proposes itself, from its home, in three steps:
244
245
 
@@ -259,6 +260,13 @@ other than `off`, or a soul `knowledge.propose` key, refuses the spawn
259
260
  that would leave the host and prints the complete outgoing diff, the
260
261
  commit text, the PR body with its provenance block and a digest; nothing
261
262
  remote changes, and the instance reads that diff once as a reviewer would.
263
+ It refuses only what protects against real harm (for example an unsafe
264
+ file, a non-Markdown file where Markdown is required, `okf-base.json`, an
265
+ invalid base or the content screen's findings);
266
+ a path outside the write surface, a claim with no subject, a broken link
267
+ or an unfilled `proposal.md` is a `W_*` warning, listed under
268
+ `## Check warnings` in the PR body, which the maintainer resolves before
269
+ merging.
262
270
  `--apply` with that digest commits, pushes a new
263
271
  `okf-proposal/<date>-<uuid>` branch (create-only, never the accepted
264
272
  branch), labels it `okf-proposal` and opens one PR per changed base. One
@@ -321,9 +329,10 @@ names but does not judge its content. A bounded content screen runs over
321
329
  everything that would leave the host before every push, comment and merge,
322
330
  and the host's `screen-terms-file` can add terms to it; a clean screen is not
323
331
  proof. The proposer reads its outgoing diff before the push; the maintainer,
324
- the second reader, reads the already-pushed pull request. The write surface,
325
- the provenance checks and the soul's opt-out are enforced in code; none is a
326
- privacy guarantee. The opt-out ([The soul's opt-out](#the-souls-opt-out)) is
332
+ the second reader, reads the already-pushed pull request. The refusals of
333
+ `--check`, the provenance checks and the soul's opt-out are enforced in code;
334
+ a path outside the write surface is only a warning the maintainer resolves.
335
+ None is a privacy guarantee. The opt-out ([The soul's opt-out](#the-souls-opt-out)) is
327
336
  per soul and covers that soul's own proposals, not what its instances write
328
337
  to other seats. A soul whose work is reading third-party correspondence
329
338
  should opt out.
@@ -9,9 +9,9 @@ or workspace membership alone does not make a package official.
9
9
 
10
10
  | package | release | capabilities | package souls |
11
11
  |---|---|---|---|
12
- | `oats.framework` | `oats-framework/v1.27.0` (this repository) | `oats.core`, `oats.setup`, `oats.onboarding`, `oats.support`, `oats.knowledge-theory`, `oats.coordination-setup` | `knowledge-theory-expert`, `oats-setup-expert` |
12
+ | `oats.framework` | `oats-framework/v1.28.0` (this repository) | `oats.core`, `oats.setup`, `oats.onboarding`, `oats.support`, `oats.knowledge-theory`, `oats.coordination-setup` | `knowledge-theory-expert`, `oats-setup-expert` |
13
13
  | `oats.okf` | `v6.3.0` | `oats.okf` (knowledge), `oats.okf-harvest`, `oats.okf-maintenance` | `knowledge-harvester`, `knowledge-maintainer` |
14
- | `oats.aweb` | `v1.46.0` | `oats.aweb` (messaging), `oats.aweb-tasks` (tasks) | |
14
+ | `oats.aweb` | `v1.47.0` | `oats.aweb` (messaging), `oats.aweb-tasks` (tasks) | |
15
15
  | `oats.engineering` | `v1.18.0` (frozen: see Coordination templates) | `oats.engineering-expert`, `oats.developer`, `oats.code-review`, `oats.maintainer` | `code-reviewer` |
16
16
  | `oats.authoring` | `v1.0.3` | `oats.authoring` | |
17
17
  | `oats.cloning` | `v1.1.0` | `oats.cloning` | `cloner` |
package/docs/packages.md CHANGED
@@ -55,7 +55,7 @@ packages:
55
55
  - **Bare version** (`v5.8.0`, `5.8.0`, `1.0.0-rc.1`): the id is looked up in
56
56
  the official catalog — `package-catalog.json` in the `oats` repo, or the file
57
57
  named by `OATS_PACKAGE_CATALOG` — which supplies the repo url, the tag
58
- convention (`v5.8.0` or `oats-framework/v1.27.0`) and the payload path. An id
58
+ convention (`v5.8.0` or `oats-framework/v1.28.0`) and the payload path. An id
59
59
  the catalog does not know is `E_PACKAGE_MISSING` ("use `git:<repo>@<ref>` for
60
60
  a package outside the catalog"). The catalog is the reviewed official list
61
61
  ([official-catalog.md](official-catalog.md)) and the only way a
@@ -78,9 +78,9 @@ members:
78
78
  - git:github.com/acme/agents
79
79
  - git:github.com/acme/platform
80
80
  packages:
81
- oats.framework: v1.27.0
81
+ oats.framework: v1.28.0
82
82
  oats.okf: v6.3.0
83
- oats.aweb: v1.46.0
83
+ oats.aweb: v1.47.0
84
84
  teams:
85
85
  platform: { team: "platform:acme.aweb.ai", description: Platform engineering }
86
86
  defaults:
@@ -155,7 +155,7 @@ Declaring a package in the workspace's `packages:` is the trust decision
155
155
  ## `oats package add | remove`
156
156
 
157
157
  ```bash
158
- oats package add oats.aweb v1.46.0 # a catalog version
158
+ oats package add oats.aweb v1.47.0 # a catalog version
159
159
  oats package add acme.tools git:github.com/acme/tools@v0.4.0
160
160
  oats package remove acme.tools
161
161
  ```
@@ -432,11 +432,11 @@ A soul that names one of the package's capabilities with
432
432
  "policy": "docs/official-catalog.md",
433
433
  "packages": {
434
434
  "oats.okf": { "url": "https://github.com/awebai/oats-okf.git", "ref": "v6.3.0", "path": "oats-package" },
435
- "oats.framework": { "url": "https://github.com/awebai/oats.git", "ref": "oats-framework/v1.27.0", "path": "oats-package" }
435
+ "oats.framework": { "url": "https://github.com/awebai/oats.git", "ref": "oats-framework/v1.28.0", "path": "oats-package" }
436
436
  }
437
437
  }
438
438
  ```
439
439
 
440
- `ref` carries the tag convention: a workspace's `oats.framework: v1.27.0`
441
- resolves to tag `oats-framework/v1.27.0`. Resolving through the catalog never
440
+ `ref` carries the tag convention: a workspace's `oats.framework: v1.28.0`
441
+ resolves to tag `oats-framework/v1.28.0`. Resolving through the catalog never
442
442
  advances a lock by itself: `oats sync` does, and says so.
@@ -0,0 +1,264 @@
1
+ # OATS 0.69.24
2
+
3
+ Unreleased
4
+
5
+ The kernel can run an agent on a session backend a capability supplies, beside
6
+ tmux (session providers, experimental until the follow-up change). The Desktop
7
+ shows an agent's board around its terminal: its facts, and its questions
8
+ answered by number. OATS now refuses to start on Node.js older than 22.
9
+ oats.framework 1.28.0 and oats.aweb 1.47.0 are pinned.
10
+
11
+ ## Upgrade
12
+
13
+ - **The catalog pins oats.framework 1.28.0 and oats.aweb 1.47.0**; oats.okf
14
+ stays 6.3.0. `oats sync` takes them. Existing homes take oats.core 2.16.0
15
+ (the `tui` default) through `oats instance upgrade` or a respawn.
16
+ - **A legacy oats.aweb grant seat** (one spawned with oats.aweb 1.12–1.24)
17
+ fills its `.oats-aweb/seat.json` once at its next start, from facts that
18
+ must agree; otherwise its start refuses with the respawn remedy. Such a
19
+ seat no longer follows a changed default team: respawn it to move it.
20
+
21
+ - **What users must do: move to Node.js 22.5 or later.** OATS now stops before
22
+ every command on Node below 22, including `version`, the reader and
23
+ capability commands. If you run OATS on Node 20 today, upgrade Node before continuing.
24
+ Node 18 and 20 receive `E_NODE_UNSUPPORTED` with exit 1, as one failure
25
+ envelope on stdout with `--json`, or a message on stderr without it. Remote
26
+ probes and Desktop discovery retain the host's refusal message. The npm
27
+ package retains its Node 22.5.0 floor so installation warns about it.
28
+
29
+ - **Respawn an instance of `oats-maintainer` or `oats-setup-admin` once**
30
+ after `oats sync`. An instance keeps the `okf.json` its spawn recorded, so
31
+ one started before this release still reads and proposes to its old node; a
32
+ new spawn reads the soul's subjects. The soul needs a knowledge base bound
33
+ with the subject layout: on a base that is not bridged, its spawn refuses
34
+ with `E_CONFIG` naming the base
35
+ ([okf.json version 2](../knowledge.md#the-souls-okfjson)).
36
+
37
+ ## Added
38
+
39
+ - **Doctor reports Git's cache support.** Text and JSON doctor output name
40
+ the Git version and whether it supports partial caches or uses the
41
+ whole-tree fallback. This uses the existing `information[]` field, adds no
42
+ warning and refuses nothing. Git 2.45 or later is recommended for partial
43
+ caches; an older Git works with whole-tree fetches.
44
+
45
+ - **The Desktop and the TUI show an agent that runs on a session provider
46
+ as OATS reports it.** A session provider is a session backend a capability
47
+ supplies, in place of tmux; this release's kernel reports them (session
48
+ providers, below), and no released provider exists yet. For such an agent:
49
+ - its state is the one OATS reports, and no client asks tmux about it;
50
+ - Start, Restart, Stop and Retire are offered by its state, as for any
51
+ agent;
52
+ - it shows as needing input whenever OATS reports that it waits on you
53
+ and it is not stopped, and the Desktop's Home lists it under Running
54
+ while it runs;
55
+ - there is no terminal to open. The clients say where it runs ("Runs in a
56
+ bb thread; open it in bb.") and show the link the provider offers: the
57
+ Desktop in the row's context menu, opened in your browser, and the TUI
58
+ printed under the line after `Enter`. Neither opens, splits or restores
59
+ a terminal tab for it;
60
+ - the row's tooltip in the Desktop names the backend and the provider's
61
+ thread id.
62
+
63
+ An agent in tmux, and one recorded on Herdr, read as before
64
+ ([how the clients read `backend`](../desktop-cli-api.md#row-backend-clients)).
65
+
66
+ - **The agent view in the Desktop.** With an OATS that reports the agent's board (feature `instance-board`), an agent's terminal tab shows what the agent keeps in your view and what it asks you:
67
+ - **Facts** above the terminal: `label · value unit`, "updated <time>", at most three shown and the rest folded, a sparkline once the view has read the history, and links shown as attributed text: a press shows the whole address, and Open in browser opens it. In a short pane the facts give way so the terminal keeps its room. × or u hides a fact on this computer only, until its value changes.
68
+ - **Questions** under the terminal, one at a time, with "question n of N", ‹ › and the dots, and a context fold (c opens it from the keyboard).
69
+ - **Answering:** The questions and the facts are F6 regions (Ctrl+Shift+F6 from the terminal reaches the questions), and the palette's Agent: questions and facts (focus) moves through them; no chord takes Alt+T, Claude Code's extended thinking. A number puts `<id>: <option>` on the agent's input line, unsent, and your Enter in the terminal sends it. No bare digit is ever typed, and nothing is typed into a terminal that isn't on screen, or into one in its scrollback or one it can't check. A question's card stays until the agent clears it. On an agent on another machine the numbers are disabled and the zone says to type the answer in its terminal. While the agent's terminal shows a harness prompt (a live `permission` or `question` claim), the numbers are disabled and the zone says to answer the prompt there first.
70
+
71
+ Without the feature the terminal tab and its waiting row are as before.
72
+
73
+ - **Session providers: a capability can supply a session backend** (feature
74
+ `session-providers`). tmux stays the kernel's own backend and the default,
75
+ and tmux homes are unchanged. A capability declares `sessionProviders` in
76
+ its manifest, four of its own commands (`launch`, `inspect`, `input`,
77
+ `stop`, and optionally `inspectMany` and `check`), and `oats spawn <soul>
78
+ --backend <name>` (a schedule's `backend`, a trigger's `spawn.backend`)
79
+ runs the instance there. OATS keeps the home, the records, the lifecycle
80
+ and the answers.
81
+ - **The session-provider wire is experimental**: it is not stable until
82
+ the follow-up change (sandbox allowances and wire 1.3), and no released
83
+ provider speaks it.
84
+ - A provider home records `backend` and `session: {provider, endpoint}`.
85
+ Its status row carries `backend`, `session: {provider}` and `open` (a
86
+ link to show instead of a terminal), and on its own host the endpoint's
87
+ thread id; the endpoint is on its local inspect, spawn and start answers
88
+ only, and nothing of it is relayed to another machine. `oats inspect` and
89
+ `oats capabilities show` list `sessionProviders`.
90
+ - A spawn holds its new home's lifecycle claim while the provider
91
+ launches, and the home's receipt records that a launch was sent until
92
+ its outcome is recorded (`runtime.launchSent`): a spawn or a start
93
+ killed mid-launch leaves a home that reads as unknown, whose next stop,
94
+ start or retire first asks the provider to stop what it started.
95
+ - Each command of a provider runs in a process group of its own, ended
96
+ whole at its timeout. A `launch`, an `input` or a `stop` whose `oats`
97
+ command was killed is ended, with its group, by the next stop, start,
98
+ input or retire of the home, before the provider is called again: it
99
+ starts no session that OATS has no record of. A stop of a provider home
100
+ waits for an input that is being sent (up to its timeout) before it
101
+ asks the provider. A provider's command runs its helpers
102
+ inside that group.
103
+ - A stop made with OATS stays stopped: inspect answers stopped and every
104
+ input is refused `E_SESSION_NOT_RUNNING` until the home is started,
105
+ without asking the provider.
106
+ - Only a lost session is resumed, in its own conversation: one stopped by
107
+ `oats host run`'s shutdown, or one that is gone while OATS did not stop
108
+ it. `oats session resume` sends it one notice through the provider's
109
+ input, under a cap of 3 resumes of one home in an hour; any input to a
110
+ lost session counts as one.
111
+ - A provider's readiness (`check`) is asked right before every launch
112
+ (`E_BACKEND_NOT_READY`), and `oats readiness` relays it.
113
+ - New codes: `E_BACKEND_UNKNOWN`, `E_BACKEND_AMBIGUOUS`,
114
+ `E_BACKEND_HARNESS`, `E_BACKEND_LAUNCH_ARG`, `E_BACKEND_NOT_READY`,
115
+ `E_SESSION_NO_TERMINAL` (attach), `E_SESSION_PROMPT_OPEN` and
116
+ `E_SESSION_INPUT_QUEUED` (input).
117
+ - A routed `--backend <provider>` is sent only to a host that lists the
118
+ feature. A kernel without it that meets a provider home only refuses.
119
+
120
+ See [Session providers](../execution-targets.md#session-providers) and, for
121
+ a capability's author,
122
+ [Session providers (`sessionProviders`)](../capabilities.md#session-providers-sessionproviders).
123
+ - **A harness-package requirement can name its backends.** A `requires` row
124
+ of a capability manifest takes `backends: ["tmux"]`, `["session-provider"]`
125
+ or both; without it the row applies to every launch. A kernel without
126
+ `session-providers` applies such a row to every launch, so a package that
127
+ uses the key raises its OATS floor.
128
+
129
+ ## Changed
130
+
131
+ - **Desktop: Home shows what needs you, and the Workspace page has its own row and a way back.**
132
+ ([Workspaces](../desktop.md#workspaces))
133
+ - **Home is one centred column.** **Needs you** comes first: each agent waiting on you, with its
134
+ soul, its machine and its message, newest first. Eight show at most, then **Show all N**. With
135
+ none, "Nothing needs you."
136
+ - **One summary line** counts the rest: "24 agents · 21 running · 3 need you · 0 stopped".
137
+ - **The workspace is one row of tiles**, each a button with its count: Agents, Souls,
138
+ Capabilities, Teams, Machines and Automations (with the red dot while a run failed). They show
139
+ without scrolling, even with eight agents waiting on you. The Graph tile is gone; Machines keeps
140
+ its List and Map.
141
+ - **Running**, under the tiles, shows five running agents, then **Show all N running**. No row
142
+ says "running" any more: the heading does.
143
+ - **The sidebar's pinned rows are Home, Workspace and Automations.** Opening any Workspace page
144
+ (from a tile, the command palette, ⌘2 / Ctrl+2, or a soul's or capability's page) marks the
145
+ Workspace row, no longer Home's. Its palette command is now **View: Workspace**.
146
+ - **Every Workspace page says where you are, and the way back:** **Home › Workspace › Souls**
147
+ above the tabs, and **Home › Workspace › Souls › dev** on a soul's page, each part a link. Esc
148
+ goes up one level: twice from a soul's page reaches Home.
149
+ - **Capabilities shows the Workspace defaults again**, above the list: the Core slots (Knowledge,
150
+ Messaging, Tasks) and the Additional defaults, each as the list's own row.
151
+ - **Soul cards read more clearly:** the name larger and bolder, what the soul does in muted body
152
+ text, its agents and kind on a smaller line.
153
+
154
+ - **Desktop: the Workspace pages look like the prototype.** ([Capabilities and spawning](../desktop.md#capabilities-and-spawning))
155
+ - **The Dark theme takes Tokyo Night's blues and greys** in the window: backgrounds, raised surfaces, ink, the blue
156
+ accent, green, amber and red. Every text colour still meets WCAG AA; the terminals keep their own colours, and their
157
+ cells their size. Light and Solarized keep their colours.
158
+ - **Souls**: **Grid** | **Table** sits beside the page's title. The first line counts the souls and those with
159
+ agents running ("You have 10 souls, 4 with agents running, …"). Group headings are small capitals with their count.
160
+ The grid has three columns at most; a card is a calmer tile, its name in semibold, what it does on two lines, then
161
+ one meta line, and a refused spawn says so in one quiet line.
162
+ - **Capabilities**: a first line says what a capability is, how many there are and how many are official, and what
163
+ ✓, **core** and **default** mean. The rows sit on the page with a thin rule under each, the ✓ or **by** right after
164
+ the name, **core** and **default** as small pills. **Workspace defaults** stays above **All capabilities**, styled the
165
+ same way.
166
+ - **A soul's page and a capability's page** use the same type and spacing: a smaller title, section titles in small
167
+ capitals, the capability and skill rows as Capabilities' rows, quieter tabs and folds.
168
+ - **The status line shows the page's keys** on these pages: "↑↓ move · Enter open · v table · Esc Home" on Souls,
169
+ and "Esc back" on a soul's or a capability's page.
170
+
171
+ - **oats.framework 1.28.0** (`oats.core` 2.16.0, `oats.coordination-setup`
172
+ 1.1.1, `oats.knowledge-theory` 1.1.3, `oats.onboarding` 1.3.1, `oats.setup`
173
+ 2.16.2).
174
+ - `oats.core` adds `tui: "default"` to an agent's Claude settings when the
175
+ key is absent, so Claude's fullscreen-renderer offer no longer takes a
176
+ wake; a value already there stays (#1692). `/oats-operate` names the Node
177
+ floor and doctor's `git-version:` line, from OATS 0.69.24.
178
+ - `oats.coordination-setup` and `oats.onboarding` teach related work: a
179
+ lead's children and its sibling leads, with `--related-work` from OATS
180
+ 0.69.23 (#2022).
181
+ - `oats.knowledge-theory`'s proposal reference names `oats.okf` 6.3's
182
+ division of work as an option (#2030).
183
+ - `oats.setup`'s `oats-agent-permissions` skill explains why launch
184
+ `args` are refused where a capability requires a harness package, and
185
+ how to choose auto mode through the launch configuration's `env` (#1692).
186
+ - Guidance text, plus the one `tui` default. Nothing changes for a
187
+ deployment beyond `oats sync` taking the new tag; existing homes take
188
+ oats.core 2.16.0 through `oats instance upgrade` or a new spawn.
189
+
190
+ - **The framework's coordination guidance teaches related work.**
191
+ `oats.coordination-setup`'s `/coordination-best-practices` and
192
+ `/adapt-to-the-organisation`, and `oats.onboarding`'s
193
+ `/oats-tailor-workspace`, say that a lead's children are spawned with
194
+ related work, and that a lead for a related effort, the first lead's
195
+ sibling, gets it too. The flag is `--related-work` from OATS 0.69.23 and
196
+ `--same-work` before, which binds a child only. Guidance text only: no
197
+ floor moves. See [Related work](../desktop-cli-api.md#related-work-oats-instance-related-work).
198
+
199
+ - **`oats-maintainer` and `oats-setup-admin` read by subject** (`okf.json`
200
+ version 2). The maintainer's subjects are `workstream/oats-direction`,
201
+ `workstream/coordination`, `system/oats-kernel`, `system/oats-desktop`,
202
+ `workstream/provider-integration`, `workstream/operations`,
203
+ `workstream/positioning` and `workstream/workspace-administration`: the
204
+ subjects that absorb the nodes it owned and read. The setup admin's are
205
+ `workstream/workspace-administration`, `workstream/operations`,
206
+ `workstream/oats-direction`, `workstream/coordination` and
207
+ `system/oats-kernel`. Their texts say each proposed claim names the subjects
208
+ it is about.
209
+
210
+ - **OATS's own coordination capabilities teach related work.** The
211
+ engineering expert's, the build and the maintainer's skills spawn with
212
+ `--related-work` where they passed `--same-work`. A lead's children
213
+ (specialists, helpers, probes, developers and reviewers) get it, as before,
214
+ and now so does a sibling lead started for an effort that arises from the
215
+ lead's own (`--relation sibling --relative-to <lead> --related-work`). An
216
+ independent lead, and a lead a maintainer launches, still don't. The kernel
217
+ and its injects set no default. An OATS before 0.69.23 refuses
218
+ `--related-work`: hosts that run these capabilities need 0.69.23 or later.
219
+ See [Related work](../desktop-cli-api.md#related-work-oats-instance-related-work).
220
+
221
+ - **Capability hooks get `OATS_BACKEND`.** Every spawn hook and launch hook
222
+ is given the session backend the instance launches on: `tmux`, or a session
223
+ provider's name.
224
+ - **A tmux home's row of `oats status --json` carries `backend: "tmux"`.** A
225
+ row with no `backend` key comes from a kernel without `session-providers`,
226
+ and reads as a tmux home's.
227
+
228
+ ## Fixed
229
+
230
+ - **Claude agent panes skip the fullscreen-renderer offer.** At spawn and
231
+ every launch, oats.core 2.16.0 adds `"tui": "default"` to a Claude home's
232
+ `.claude/settings.json` if no `tui` key is present. Any existing value is
233
+ kept. Agent panes use Claude's classic renderer even where Claude now
234
+ defaults to fullscreen, and even over a user-level `"tui": "fullscreen"`;
235
+ this keeps tmux scrollback and the renderer used by the launch-prompt
236
+ captures. An existing project `tui` still wins. If oats.core stops adding
237
+ the key, existing homes keep it until respawn. See
238
+ [oats.core](../capabilities.md#oatscore-needs-input).
239
+
240
+ ## Documentation
241
+
242
+ - **Launch args guidance matches the package-probe refusal.** Any non-empty
243
+ launch-configuration `args` is refused at spawn with
244
+ `E_LAUNCH_PROBE_UNSUPPORTED` while an active capability requires a package
245
+ for that harness; with oats.aweb, every Claude launch has such a requirement.
246
+ `claude plugin list --json` cannot carry those args, and `--settings` can
247
+ change which plugins load. The supported routes are user `settings.json`
248
+ in the Claude config selected by `CLAUDE_CONFIG_DIR` through `env`, which
249
+ the probe carries, or a wrapper executable. `/oats-agent-permissions` now
250
+ teaches `permissions.defaultMode: "auto"` in that user file, replacing its
251
+ launch-args advice. See [launch configurations](../configuration.md#launch-configurations).
252
+
253
+ - **Knowledge proposals as oats.okf 6.3 runs them.** The
254
+ [operator guide](../knowledge.md#the-flow) still said a claim
255
+ goes only in the soul's owned nodes, that the proposer settles duplicates
256
+ before proposing, and that the write surface is enforced. With 6.3 a quick
257
+ search is enough, duplicates, the canonical home and the subject are the
258
+ knowledge maintainer's, and a path outside the write surface is a `W_*`
259
+ warning under `## Check warnings` in the PR body, which the maintainer
260
+ resolves before merging. The authors' guide's
261
+ [judgment and proposal](../knowledge-reference/proposal.md#judgment-procedure)
262
+ page now says a capability may divide the work that way; the exclusions and
263
+ the screen before delivery stay with the proposer. The
264
+ `oats.knowledge-theory` copy changes with it.