instar 1.3.345 → 1.3.346

package/src/data/builtin-manifest.json

@@ -1,8 +1,8 @@
 {
   "$schema": "./builtin-manifest.schema.json",
   "schemaVersion": 1,
-  "generatedAt": "2026-06-06T05:35:52.668Z",
-  "instarVersion": "1.3.345",
+  "generatedAt": "2026-06-06T05:44:57.394Z",
+  "instarVersion": "1.3.346",
   "entryCount": 199,
   "entries": {
     "hook:session-start": {
@@ -11,7 +11,7 @@
       "domain": "identity",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/session-start.sh",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:dangerous-command-guard": {
@@ -20,7 +20,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/dangerous-command-guard.sh",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:grounding-before-messaging": {
@@ -29,7 +29,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/grounding-before-messaging.sh",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:compaction-recovery": {
@@ -38,7 +38,7 @@
       "domain": "identity",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/compaction-recovery.sh",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:external-operation-gate": {
@@ -47,7 +47,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/external-operation-gate.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:deferral-detector": {
@@ -56,7 +56,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/deferral-detector.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:self-stop-guard": {
@@ -65,7 +65,7 @@
       "domain": "coherence",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/self-stop-guard.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:post-action-reflection": {
@@ -74,7 +74,7 @@
       "domain": "evolution",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/post-action-reflection.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:external-communication-guard": {
@@ -83,7 +83,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/external-communication-guard.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:scope-coherence-collector": {
@@ -92,7 +92,7 @@
       "domain": "coherence",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/scope-coherence-collector.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:scope-coherence-checkpoint": {
@@ -101,7 +101,7 @@
       "domain": "coherence",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/scope-coherence-checkpoint.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:free-text-guard": {
@@ -110,7 +110,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/free-text-guard.sh",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:claim-intercept": {
@@ -119,7 +119,7 @@
       "domain": "coherence",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/claim-intercept.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:claim-intercept-response": {
@@ -128,7 +128,7 @@
       "domain": "coherence",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/claim-intercept-response.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:stop-gate-router": {
@@ -137,7 +137,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/stop-gate-router.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "hook:auto-approve-permissions": {
@@ -146,7 +146,7 @@
       "domain": "safety",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
       "installedPath": ".instar/hooks/instar/auto-approve-permissions.js",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "job:health-check": {
@@ -1538,7 +1538,7 @@
       "type": "subsystem",
       "domain": "updates",
       "sourcePath": "src/core/PostUpdateMigrator.ts",
-      "contentHash": "6449c8a2aaab14962a8e871a700feea312280a30e08d40f35d6c4af7b19a6e90",
+      "contentHash": "2562b5b7c37deb568e00eefebc9acc07571a5e5ccc251f6e424e7e63039ee297",
       "since": "2025-01-01"
     },
     "subsystem:scheduler": {

package/upgrades/1.3.346.md

@@ -0,0 +1,57 @@
+# Upgrade Guide — vNEXT
+
+<!-- assembled-by: assemble-next-md -->
+<!-- bump: patch -->
+
+## What Changed
+
+Honest turn-receipts — the standby (🔭) system now tells the truth about WHY a
+turn failed instead of saying "actively working" while a session is dead.
+
+Three incidents in two days had one symptom: a delivered message that never got
+a reply, while the standby showed "🔭 actively working." Root: the tier-3
+assessment classified a session "working" whenever it had a live child process
+— but a rate-limited, policy-wedged, or context-exhausted session HAS a live
+process (the model CLI is alive, just failing every turn). The live process
+forced the lie. Separately, the "conversation too long" standby fired on that
+phrase anywhere in the buffer, so a stale scrolled-past mention surfaced as
+noise on healthy sessions.
+
+New `StuckSignatureClassifier` (pure, tail-gated, signal-only) classifies a
+live-but-failing session from its LIVE tmux tail — rate-limited / policy-wedge /
+context-wedge / context-too-long — and PresenceProxy surfaces the real reason
+instead of "working". Tail-gating (the signature must be the live tail, not a
+scrollback mention) is the same discriminator that kills the "conversation too
+long" noise. The classifier defers to any recovery sentinel that already owns a
+session's recovery, so the user always hears one voice. Recovery itself is
+unchanged — this only makes the messaging honest.
+
+## What to Tell Your User
+
+When a message of yours goes unanswered, I'll now tell you the real reason if I
+can't reply — "I've hit the usage limit, resets 10:30pm", "my session got stuck
+on a content-policy error, resend your last message" — instead of a misleading
+"actively working." And the "conversation too long" messages that used to pop up
+when nothing was wrong are gone: that only fires now when it's actually
+happening.
+
+## Summary of New Capabilities
+
+- `StuckSignatureClassifier` — tail-gated classification of a live-but-failing
+  session (rate-limited / policy-wedge / context-wedge / context-too-long) with
+  an honest user-facing message per kind. Pure + signal-only.
+- PresenceProxy tier-3 surfaces the honest reason instead of "working", defers
+  to an owning recovery sentinel (`isStuckRecoveryActive`), and no longer fires
+  the context-too-long notice on a stale scrollback mention.
+- CLAUDE.md "Honest standby (turn-receipts)" section so agents can explain the
+  new behavior and the noise fix.
+
+## Evidence
+
+Grounded in the real 2026-06-04/05 incidents (rate-limit, AUP-wedge) using the
+verbatim pane text. 30+ new unit tests both sides of every boundary
+(honest-when-stuck AND quiet-when-fine, incl. the stale-scrollback noise case
+and prose-mentions-a-limit false positive) + a behavioral test driving the real
+fireTier3 with a live child process (wedge → honest, NOT "working"; deference →
+silent) + migrator add/idempotent + server.ts wiring guard. All 119 existing
+presence tests pass; tsc clean; preflight PASS.

package/upgrades/honest-turn-receipts.eli16.md

@@ -0,0 +1,47 @@
+# ELI16 — Honest Turn-Receipts
+
+## The problem
+
+Three times in two days, the same thing happened: you sent me a message, you
+saw "✓ Delivered" and "🔭 actively working" — and then nothing. No reply. You
+found out by screenshot each time. Three different reasons underneath (I'd hit
+the usage limit, or my session got stuck on a content-policy error, or the
+conversation ran out of room), but ONE misleading symptom: the system kept
+telling you I was "actively working" when I was actually dead in the water.
+
+Why did it lie? Because the way it checked "is this session working?" was "is
+its program still running?" — and a stuck session's program IS still running.
+It's alive, it's even printing stuff to the screen; it just can't answer. So
+the live program fooled the check into saying "working."
+
+You also noticed a related annoyance: "conversation too long" messages popping
+up often even when the conversation was fine. Same kind of bug — the system was
+finding that phrase ANYWHERE on the screen, including an old one from an hour
+ago that had already been dealt with and scrolled up out of the way.
+
+## What this fixes
+
+Now, before the system says "working," it reads the bottom of the screen — the
+LIVE part, what's happening right now — and checks for the known "alive but
+can't reply" situations. If it finds one, it tells you the truth instead:
+
+- "I've hit the usage limit (resets 10:30pm) — I'll pick back up automatically,
+  your messages aren't lost."
+- "My session got stuck on a content-policy error — please resend your last
+  message."
+- "This conversation got too long — I'm starting fresh, resend your last one."
+
+And the "conversation too long" noise is gone: it only fires if that's what's
+ACTUALLY happening right now at the bottom of the screen, not an old mention
+that scrolled by.
+
+## Two safety rules
+
+1. **It only reads — it doesn't act.** Fixing the stuck session is still the
+   job of the recovery system. This part just stops lying about what's wrong.
+2. **One voice.** If the recovery system is already telling you about a stuck
+   session, this stays quiet so you don't get two messages about the same thing.
+
+Tested with the exact text from the real screenshots you sent — 30+ new checks
+covering both "say the honest thing when stuck" and "stay quiet when fine,"
+plus all 119 existing standby checks still pass.

package/upgrades/side-effects/honest-turn-receipts.md

@@ -0,0 +1,90 @@
+# Side-Effects Review — Honest Turn-Receipts
+
+**Version / slug:** `honest-turn-receipts`
+**Date:** `2026-06-06`
+**Author:** `Echo (instar-dev agent, session-robustness topic — Justin: "finish out item number four" + the "conversation too long is noise" observation)`
+**Second-pass reviewer:** `self-adversarial pass over false-positive surface (the one way an honest classifier can become a new liar) + double-messaging`
+
+## Summary of the change
+
+A pure tail-gated `StuckSignatureClassifier` recognizes a live-but-failing
+session (rate-limited / policy-wedge / context-wedge / context-too-long) from
+its LIVE tmux tail and returns an honest user-facing message. PresenceProxy's
+`fireTier3` runs it after the existing quota check and before the process-tree
+"working" assessment, surfacing the real reason instead of "working", deferring
+to an owning recovery sentinel, and preserving the context-exhaustion recovery
+path. The un-tail-gated context-exhaustion block it replaces was the source of
+the "conversation too long" noise. Files: `StuckSignatureClassifier.ts` (new),
+`PresenceProxy.ts`, `server.ts` wiring, `PostUpdateMigrator.ts` section, tests.
+
+## Decision-point inventory
+
+- Stuck classification — **add (detector)** — signal only; tail-gated.
+- fireTier3 placement (after quota, before process-tree) — **modify** — the
+  exact seam where a live process forced the "working" lie.
+- context-exhaustion block — **replace** — same recovery path, now tail-gated
+  (fixes the noise) and folded into the unified honest block.
+- `isStuckRecoveryActive` deference — **add (suppression)** — net reduction in
+  messaging authority (yields to an owning sentinel).
+
+## 1. Over-block
+
+Nothing is blocked — no authority added. The honest message can only fire at
+tier-3 (5 minutes of an unanswered user message), so a session that replied
+promptly never reaches it. The deference callback REMOVES a double-message case
+(standby + sentinel both speaking).
+
+## 2. Under-block (false-positive surface — the real risk)
+
+The danger of an honest classifier is becoming a NEW liar (saying "stuck" about
+a healthy session). Defenses, each tested:
+- **Tail-gating:** the signature must be the live tail, not scrollback — kills
+  the stale "conversation too long" mention (the reported noise) and a session
+  that quoted an error then kept working.
+- **Prose-vs-block rate-limit patterns:** "you've hit your limit" / "limit ·
+  resets" match; "when you hit your usage limit, the session pauses" does not.
+- **AUP repetition gate** (inherited from classifyWedgeTail): a single benign
+  policy rejection is not a wedge.
+- **Normal-compaction suppression:** the compaction lifecycle banner is not
+  treated as context-too-long.
+Residual under-block: a stuck session whose signature the provider rewords
+escapes — inherent to signature matching; the tier-3 LLM path still runs as the
+fallback for the unmatched case (unchanged).
+
+## 3. Level-of-abstraction fit
+
+The classifier is a pure module beside the sentinels whose detectors it reuses
+(`classifyWedgeTail`). The wiring lives in `fireTier3` (which owns the
+assessment) and reuses the existing `sendProxyMessage` surface and the composed
+`wedgeRecoveryActive` checker (same one the SessionReaper veto uses). No new
+message path, no new state store.
+
+## 4. Signal vs authority compliance
+
+**Required reference:** `docs/signal-vs-authority.md`
+
+- [x] Pure signal. The classifier decides nothing; it answers a question.
+  Recovery is unchanged (the sentinels'). The only behavioral delta is WHICH
+  honest string the standby sends, plus a new SUPPRESSION (deference). No new
+  decision-maker, and a net reduction in messaging authority.
+
+## 5. Interactions
+
+- **Quota-exhaustion block (runs first):** owns the bare usage-limit form; the
+  honest classifier is the backstop for the forms it misses + the wedges +
+  tail-gated context-too-long. No double-handling (the first to match returns).
+- **RateLimitSentinel:** already suppresses ALL tiers via
+  `hasActiveRateLimitRecovery` (unchanged); the honest rate-limit branch is the
+  backstop when no sentinel owns it (e.g. the cross-machine EXO case).
+- **ContextWedgeSentinel:** when it is mid-recovery, `isStuckRecoveryActive`
+  makes the standby defer; otherwise the honest message is the only voice.
+- **Existing tests:** all 119 presence-proxy tests pass unmodified; the
+  context-exhaustion test's source-grep section was updated to the new wiring.
+
+## 6. External surfaces / 7. Rollback
+
+No API, no route, no config key (default-on behavioral change; the new config
+field is an optional injected callback wired in server.ts). One idempotent
+CLAUDE.md section (marker 'Honest standby (turn-receipts)'). Rollback = revert;
+the standby returns to classifying live-but-failing sessions as "working" and
+the context-too-long noise returns.