axstack 0.19.1 → 0.20.1

package/README.md

@@ -79,7 +79,7 @@ targets are reported without normal-path adoption. Install exits nonzero when
 an instruction conflict is preserved, while clean and idempotent installs exit
 successfully.
 
-The public bundle preserves three canonical 25-role inputs:
+The public bundle preserves three canonical 24-role inputs:
 [mixed](profiles/presets/mixed.json),
 [codex-only](profiles/presets/codex-only.json), and
 [claude-only](profiles/presets/claude-only.json). Each is exactly
@@ -114,11 +114,15 @@ answer trust or permission prompts on the worker's behalf, and never create a
 duplicate writer. A `worker_done` advances work only when its Task and Dispatch
 match the active attempt and its revision evidence verifies.
 
-The user lifted the native-watch hold by decision on 2026-09-16. The driver
-every 15 minutes dispatches and exits; the watchdog is model-free and
-read-only, has no gate, and records `watchdog.log`; there is no watch deadline
-for automations. Axstack uses no historical fallback and introduces no custom
-scheduler.
+Two logical native PR-manager lanes run on staggered 15-minute schedules. Each
+pass uses a fresh finite session in its persistent dedicated workspace. Review
+and watch each admit at most five bounded PR jobs;
+waiting PRs stay covered without reserving slots. Axstack adds no custom
+scheduler, queue engine, or decision interpreter. Sessions reconcile before
+admission, save durable continuity and decisions, clean only positively
+identified owned unused setup shells, and self-close as their final action.
+Native fresh-session, overlapping-pass, recovery, and VPS resource behavior
+require a canary before activation.
 
 Mobile completion/reply behavior remains unverified. Structural checks and
 qualitative scenario evaluation are not live runtime proof.

package/docs/installation.md

@@ -51,7 +51,7 @@ profiles/presets/codex-only.json
 profiles/presets/claude-only.json
 ```
 
-Each has exactly `{ "version": 1, "roles": [...] }` with the same 25 stable
+Each has exactly `{ "version": 1, "roles": [...] }` with the same 24 stable
 role IDs. Installation writes `<skills-dir>/axstack/roles.json` as
 `{ "version": 1, "preset": "<selected preset>", "roles": [...] }` and records
 its ownership hash like every other installed skill asset. There is no second
@@ -150,7 +150,7 @@ to rewrite them.
 ## Role behavior after installation
 
 The runtime reads `roles.json` from the installed shared root `skills/axstack/`.
-A new run records the selected preset plus all 25 role rows. An active run keeps
+A new run records the selected preset plus all 24 role rows. An active run keeps
 that snapshot after a later preset install unless the user explicitly changes
 it and accepts the resulting evidence invalidation.
 

package/docs/workflows.md

@@ -40,7 +40,7 @@ only affected work.
 
 Installation requires one explicit canonical preset. The three bundle files
 under `profiles/presets/` each contain exactly
-`{ "version": 1, "roles": [...] }` and the same 25 stable IDs.
+`{ "version": 1, "roles": [...] }` and the same 24 stable IDs.
 
 The current chat drives on whatever model runs it; no preset carries a driver
 role.
@@ -139,10 +139,10 @@ session and evidence remain valid.
   authorized-maintenance scope. A changed head or comment is an event, not
   repair authority. Repairs return to the original author only for a
   run-launched session and receive refreshed authored review before scoped
-  `gh stack` publication. `gh stack` publication does not apply to automation
-  repairs: the automation session (or a dispatched `axstack-author`) repairs in
-  a per-PR child worktree, the local SHA is reviewed and gated, and the repair
-  lands by fast-forward `git push` after `proceed`.
+  `gh stack` publication. `gh stack` publication does not apply to manager
+  repairs: the bounded PR job repairs in its per-PR child worktree, the local
+  SHA receives the actual-author-provenance review, and the exact candidate
+  lands by fast-forward `git push` after final readback.
 - `axstack-audit` separates execution outcome, procedure, and measurement
   coverage with evidenced denominators; it proposes but never self-edits.
 
@@ -173,40 +173,45 @@ gate. An applicable `Notification policy` may use `axstack-relay`; otherwise the
 current Orca conversation is the fallback. The relay normally delivers one-way
 through native `hermes send`: it checks CLI lookup and the configured target,
 binds the recipient, deduplicates on the run record, and records the returned
-`message_id`. The PR automation's decision tokens are the narrow exception: a
-fixed Hermes script writes the user's bound decision to a file for the driver
-to consume; no session polls Telegram. Delivery failure never clears the
-underlying hold.
+`message_id`. PR-manager notifications point the user to GitHub or a durable
+user-owned conversation; Telegram delivery, replies, and silence grant no action
+authority. Delivery failure never clears the underlying hold.
 
-Healthy watch observations remain quiet. The optional `axstack-monitor` is
-read-only and never sends; `axstack-watchdog` never mutates GitHub. Under the
-rev-3 PR automation it performs four model-free health checks and sends new
-occurrences directly, with no health gate.
+Healthy watch observations remain quiet. The optional `axstack-monitor` is a
+read-only observer for standalone watches and never sends.
 
 ## Native watch automations
 
-The accepted PR-automation contract is a 15-minute driver automation and an
-hourly watchdog, with quiet healthy checks, deduplicated occurrences, one live
-dispatch marker per PR, and decision tokens for user-authorized actions.
+The accepted contract has exactly two logical native manager lanes: review at
+minutes `0,15,30,45` and watch at `7,22,37,52`. Each scheduled pass uses a
+fresh finite session in a persistent dedicated workspace, scans complete
+discovery pages, and admits at most five executing
+PR jobs. Waiting PRs stay covered and consume no slot after owned descendants
+settle. Each job uses one repository-parented worktree; managers never check out
+PR branches in their own workspaces.
+
+Every pass reconciles saved, GitHub, and native Orca state before admission. A
+confirmed same-lane manager makes the new duplicate do no work or shared-record
+write and close only itself. Normal teardown settles descendants, saves durable
+continuity and decisions, closes only positively identified owned unused setup
+shell terminals, then self-closes as the final action. It never blanket-closes
+the workspace or cleans preserved candidates, evidence, user sessions, unknown
+liveness, `user_takeover`, or ambiguous publication state.
 
-The driver is the automation session itself, with no `axstack-monitor` or
-`axstack-owner` role row. It uses the agent selected in Orca; changing that
-selection does not change the configured reviewers or impose a model hold.
 Requested reviews cover any accessible repository; automatic repairs retain
-their separate explicit repository scope. Orca owns scheduling and run history;
-Axstack adds no custom scheduler, polling loop, or historical runtime fallback.
+their `defi-com/monorepo` and `defi-com/mobile` scope. Orca owns schedules,
+sessions, Tasks, and Dispatches. Axstack adds no custom scheduler, queue engine,
+cursor files, polling loop, or historical runtime fallback.
 
 ## Automations
 
-Two native Orca automations run the installed skills: a driver every 15 minutes
-that discovers work through four GitHub searches, dispatches exact-head peer
-reviews and own-PR repairs, consumes decision tokens, and exits; and an hourly
-model-free watchdog that evaluates four liveness checks and exits without
-launching a session. PR reviewers use exactly three escalation criteria. A gate
-`escalate` opens a bound decision token and exits, while `proceed` permits only
-the verdict or fast-forward push supported by the reviewed evidence. There are
-no `COMMENT` reviews, obligations, watch deadline, terminal cleanup sweep, or
-health gate. The current operational contract is
+The review and watch managers use short packaged prompts that load the current
+relative contract and invoke `axstack-review` or `axstack-watch`. Bounded jobs
+publish ordinary exact-head review verdicts or reviewed fast-forward repairs;
+the human merges. Exceptional security, permanent-on-chain, or architectural
+decisions remain actionable in GitHub or a durable user-owned conversation
+after manager self-close, with an authorized deduplicated Telegram notification.
+The current operational contract is
 `skills/axstack/references/automations.md`.
 
 ## Run record and evidence

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "axstack",
-  "version": "0.19.1",
+  "version": "0.20.1",
   "description": "Axstack installer and setup CLI: installs owned chat skills and role data, configures supported harness settings, and checks Orca capabilities.",
   "keywords": [
     "claude-code",

package/profiles/presets/claude-only.json

@@ -152,16 +152,7 @@
       "model": "claude-sonnet-5",
       "modeId": "bypassPermissions",
       "thinkingOptionId": "low",
-      "notes": "Optional independent read-only PR observer, not the driver: the five-minute driver is the automation session itself. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, never authors, reviews, or replies. Healthy snapshots stay quiet. Hold was lifted by user decision on 2026-09-16."
-    },
-    {
-      "id": "axstack-watchdog",
-      "name": "Axstack watchdog",
-      "provider": "claude",
-      "model": "claude-sonnet-5",
-      "modeId": "bypassPermissions",
-      "thinkingOptionId": "low",
-      "notes": "Hourly independent read-only automation health checker: compares the recorded driver model with the expected one, checks run history, precheck errors, duplicate handling, and relay receipts, and passes findings to the gate under the automation-health criterion. Never mutates GitHub and never authors, reviews, or replies; its only gate-authorized automation-health send is recorded in watchdog.json."
+      "notes": "Optional independent read-only observer for a standalone PR watch. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, authors, reviews, replies, or acts as either reusable PR manager. Healthy snapshots stay quiet."
     },
     {
       "id": "axstack-auditor",

package/profiles/presets/codex-only.json

@@ -152,16 +152,7 @@
       "model": "gpt-5.6-terra",
       "modeId": "full-access",
       "thinkingOptionId": "low",
-      "notes": "Optional independent read-only PR observer, not the driver: the five-minute driver is the automation session itself. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, never authors, reviews, or replies. Healthy snapshots stay quiet. Hold was lifted by user decision on 2026-09-16."
-    },
-    {
-      "id": "axstack-watchdog",
-      "name": "Axstack watchdog",
-      "provider": "codex",
-      "model": "gpt-5.6-terra",
-      "modeId": "full-access",
-      "thinkingOptionId": "low",
-      "notes": "Hourly independent read-only automation health checker: compares the recorded driver model with the expected one, checks run history, precheck errors, duplicate handling, and relay receipts, and passes findings to the gate under the automation-health criterion. Never mutates GitHub and never authors, reviews, or replies; its only gate-authorized automation-health send is recorded in watchdog.json."
+      "notes": "Optional independent read-only observer for a standalone PR watch. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, authors, reviews, replies, or acts as either reusable PR manager. Healthy snapshots stay quiet."
     },
     {
       "id": "axstack-auditor",

package/profiles/presets/mixed.json

@@ -152,16 +152,7 @@
       "model": "claude-opus-5",
       "modeId": "bypassPermissions",
       "thinkingOptionId": "medium",
-      "notes": "Optional independent read-only PR observer, not the driver: the five-minute driver is the automation session itself. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, never authors, reviews, or replies. Healthy snapshots stay quiet. Hold was lifted by user decision on 2026-09-16."
-    },
-    {
-      "id": "axstack-watchdog",
-      "name": "Axstack watchdog",
-      "provider": "claude",
-      "model": "claude-opus-5",
-      "modeId": "bypassPermissions",
-      "thinkingOptionId": "medium",
-      "notes": "Hourly independent read-only automation health checker: compares the recorded driver model with the expected one, checks run history, precheck errors, duplicate handling, and relay receipts, and passes findings to the gate under the automation-health criterion. Never mutates GitHub and never authors, reviews, or replies; its only gate-authorized automation-health send is recorded in watchdog.json."
+      "notes": "Optional independent read-only observer for a standalone PR watch. Reads GitHub, feedback, and checks, persists event IDs, and wakes the owner only for a new actionable event. Never sends, authors, reviews, replies, or acts as either reusable PR manager. Healthy snapshots stay quiet."
     },
     {
       "id": "axstack-auditor",