instar 1.2.56 → 1.2.58

package/.claude/skills/autonomous/SKILL.md

@@ -162,7 +162,16 @@ The stop hook will catch every attempt to exit and feed your task list back. Eac
 
 ## Step 4: Completion
 
-When ALL tasks are genuinely done:
+**Preferred: a verifiable completion CONDITION (independent judge, like /goal).**
+Pass `--completion-condition "<measurable end-state>"` when starting (e.g. "all tests in
+test/auth pass and `npm test` exits 0"). Each turn, an INDEPENDENT model judges the condition
+against what you've SURFACED in the conversation — so *run the real checks and show the
+evidence in your output*. When the judge confirms it, the hook exits automatically. You do not
+self-declare done. If the judge can't be reached, the run keeps going (fail-safe). This mirrors
+the framework `/goal` feature and is harder to fool than a self-declared promise.
+
+**Legacy fallback: self-declared promise.** If no condition is set, when ALL tasks are
+genuinely done:
 
 1. Verify every task is complete (re-read the list)
 2. Run `npx tsc --noEmit` — zero errors
@@ -171,7 +180,7 @@ When ALL tasks are genuinely done:
 5. Send final report via messaging
 6. Output: `<promise>ALL_TASKS_COMPLETE</promise>`
 
-The stop hook will detect the promise and allow exit.
+The stop hook detects the promise and allows exit.
 
 ---
 

package/.claude/skills/autonomous/hooks/autonomous-stop-hook.sh

@@ -145,6 +145,7 @@ ITERATION=$(fm_get iteration)
 DURATION_SECONDS=$(fm_get duration_seconds)
 STARTED_AT=$(fm_get started_at)
 COMPLETION_PROMISE=$(fm_get completion_promise)
+COMPLETION_CONDITION=$(fm_get completion_condition)
 
 # Validate recorded session_id is a real UUID. Claude sometimes writes a custom
 # string instead of $CLAUDE_CODE_SESSION_ID; non-UUID values are treated as
@@ -267,7 +268,37 @@ if [[ -f ".instar/autonomous-emergency-stop" ]]; then
   exit 0
 fi
 
-# Completion promise (genuine completion)
+# Completion CONDITION — independent evaluator (mirrors /goal). Authoritative when
+# set; the self-declared promise below is the legacy fallback. FAIL-SAFE: if the
+# evaluator is unreachable or unsure, we keep working — never a false "done".
+EVAL_REASON=""
+if [[ -n "$COMPLETION_CONDITION" ]] && [[ -n "$TRANSCRIPT_PATH" ]] && [[ -f "$TRANSCRIPT_PATH" ]]; then
+  EVAL_MET=""
+  if [[ -n "${INSTAR_HOOK_EVAL_OVERRIDE:-}" ]]; then
+    # Test seam: "met" | "not-met" short-circuits the live evaluator call.
+    [[ "$INSTAR_HOOK_EVAL_OVERRIDE" == "met" ]] && EVAL_MET="true"
+    EVAL_REASON="override:$INSTAR_HOOK_EVAL_OVERRIDE"
+  else
+    EVAL_TAIL=$(grep '"role":"assistant"' "$TRANSCRIPT_PATH" 2>/dev/null | tail -6 \
+      | jq -r '.message.content | map(select(.type=="text")) | map(.text) | join("\n")' 2>/dev/null \
+      | tail -c 8000 || echo "")
+    EVAL_PORT=$(python3 -c "import json;print(json.load(open('.instar/config.json')).get('port',4040))" 2>/dev/null || echo 4040)
+    EVAL_AUTH=$(python3 -c "import json;print(json.load(open('.instar/config.json')).get('authToken',''))" 2>/dev/null || echo "")
+    EVAL_RESP=$(jq -nc --arg c "$COMPLETION_CONDITION" --arg t "$EVAL_TAIL" '{condition:$c,transcriptTail:$t}' \
+      | curl -s -m 35 -H "Authorization: Bearer $EVAL_AUTH" -H 'Content-Type: application/json' \
+        --data-binary @- "http://localhost:${EVAL_PORT}/autonomous/evaluate-completion" 2>/dev/null || echo "")
+    EVAL_MET=$(printf '%s' "$EVAL_RESP" | jq -r '.met // empty' 2>/dev/null || echo "")
+    EVAL_REASON=$(printf '%s' "$EVAL_RESP" | jq -r '.reason // empty' 2>/dev/null || echo "")
+  fi
+  if [[ "$EVAL_MET" == "true" ]]; then
+    echo "✅ Autonomous mode: completion condition met (independent evaluator): ${EVAL_REASON}"
+    rm -f "$STATE_FILE"
+    exit 0
+  fi
+  # Not met / unreachable → keep working; EVAL_REASON (if any) becomes next-turn guidance.
+fi
+
+# Completion promise (genuine completion — legacy/self-declared fallback)
 if [[ -n "$TRANSCRIPT_PATH" ]] && [[ -f "$TRANSCRIPT_PATH" ]]; then
   LAST_LINE=$(grep '"role":"assistant"' "$TRANSCRIPT_PATH" 2>/dev/null | tail -1 || echo "")
   if [[ -n "$LAST_LINE" ]]; then
@@ -393,7 +424,16 @@ else
   TIME_MSG="no time limit"
 fi
 
-SYSTEM_MSG="🔄 Autonomous iteration $NEXT_ITERATION ($TIME_MSG) | Complete ALL tasks, then output <promise>$COMPLETION_PROMISE</promise> | Do NOT defer to future self — if you can do it now, DO IT NOW${REPORT_DIRECTIVE}"
+# When a completion CONDITION is set, an independent judge decides "done" — steer
+# toward the condition + feed back the judge's latest reason (mirrors /goal). When
+# only a legacy promise is set, keep the self-declared-promise directive.
+if [[ -n "$COMPLETION_CONDITION" ]]; then
+  GUIDANCE=""
+  [[ -n "$EVAL_REASON" ]] && GUIDANCE=" | Not done yet: ${EVAL_REASON}"
+  SYSTEM_MSG="🔄 Autonomous iteration $NEXT_ITERATION ($TIME_MSG) | Keep working until this is TRUE: ${COMPLETION_CONDITION}${GUIDANCE} | An independent check decides done from what you SURFACE — run the real checks and show the evidence. Do NOT defer — do it now${REPORT_DIRECTIVE}"
+else
+  SYSTEM_MSG="🔄 Autonomous iteration $NEXT_ITERATION ($TIME_MSG) | Complete ALL tasks, then output <promise>$COMPLETION_PROMISE</promise> | Do NOT defer to future self — if you can do it now, DO IT NOW${REPORT_DIRECTIVE}"
+fi
 
 # Block exit and feed prompt back
 jq -n \

package/.claude/skills/autonomous/scripts/setup-autonomous.sh

@@ -14,6 +14,7 @@ REPORT_CHANNEL="telegram"   # channel that owns this job; recovery note routes h
 LEVEL_UP="false"
 TASKS=""
 COMPLETION_PROMISE=""
+COMPLETION_CONDITION=""   # verifiable end-state; an independent judge decides "done" (mirrors /goal). Preferred over the self-declared promise.
 REPORT_INTERVAL="30m"
 
 while [[ $# -gt 0 ]]; do
@@ -46,6 +47,10 @@ while [[ $# -gt 0 ]]; do
       COMPLETION_PROMISE="$2"
       shift 2
       ;;
+    --completion-condition)
+      COMPLETION_CONDITION="$2"
+      shift 2
+      ;;
     --report-interval)
       REPORT_INTERVAL="$2"
       shift 2
@@ -153,6 +158,7 @@ report_interval: "$REPORT_INTERVAL"
 last_report_at: ""
 level_up: $LEVEL_UP
 completion_promise: "$COMPLETION_PROMISE"
+completion_condition: "$COMPLETION_CONDITION"
 ---
 
 # Autonomous Session

package/dist/commands/server.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"server.d.ts","sourceRoot":"","sources":["../../src/commands/server.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAqQH,UAAU,YAAY;IACpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB,GAAG,CAAC,EAAE,MAAM,CAAC;IACb;2DACuD;IACvD,QAAQ,CAAC,EAAE,OAAO,CAAC;CACpB;AAiqDD,wBAAsB,WAAW,CAAC,OAAO,EAAE,YAAY,GAAG,OAAO,CAAC,IAAI,CAAC,CA0xLtE;AAED,wBAAsB,UAAU,CAAC,OAAO,EAAE;IAAE,GAAG,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CAsDzE;AAED;;;;;;;;;;;GAWG;AACH,wBAAsB,aAAa,CAAC,OAAO,EAAE;IAAE,GAAG,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CAuD5E"}
+{"version":3,"file":"server.d.ts","sourceRoot":"","sources":["../../src/commands/server.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AAsQH,UAAU,YAAY;IACpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB,GAAG,CAAC,EAAE,MAAM,CAAC;IACb;2DACuD;IACvD,QAAQ,CAAC,EAAE,OAAO,CAAC;CACpB;AAiqDD,wBAAsB,WAAW,CAAC,OAAO,EAAE,YAAY,GAAG,OAAO,CAAC,IAAI,CAAC,CA0zLtE;AAED,wBAAsB,UAAU,CAAC,OAAO,EAAE;IAAE,GAAG,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CAsDzE;AAED;;;;;;;;;;;GAWG;AACH,wBAAsB,aAAa,CAAC,OAAO,EAAE;IAAE,GAAG,CAAC,EAAE,MAAM,CAAA;CAAE,GAAG,OAAO,CAAC,IAAI,CAAC,CAuD5E"}

package/dist/commands/server.js

@@ -91,6 +91,7 @@ import { pickupGitSyncMessages } from '../messaging/GitSyncTransport.js';
 import { DeliveryRetryManager } from '../messaging/DeliveryRetryManager.js';
 import { SpawnRequestManager } from '../messaging/SpawnRequestManager.js';
 import { ThreadlineRouter } from '../threadline/ThreadlineRouter.js';
+import { resolveThreadlineMcpEntry } from '../threadline/mcpEntry.js';
 import { ThreadResumeMap } from '../threadline/ThreadResumeMap.js';
 import { ListenerSessionManager } from '../threadline/ListenerSessionManager.js';
 import { SystemReviewer } from '../monitoring/SystemReviewer.js';
@@ -2116,7 +2117,17 @@ export async function startServer(options) {
                 console.log(pc.yellow(`  Git sync setup: ${err instanceof Error ? err.message : String(err)}`));
             }
         }
-        const sessionManager = new SessionManager(config.sessions, state);
+        // Per-agent Codex threadline MCP override. Codex reads a SHARED
+        // ~/.codex/config.toml whose [mcp_servers."threadline"] is last-writer-wins
+        // across every codex agent on the machine — so a codex worker could load a
+        // DIFFERENT agent's threadline identity and its threadline_send would be
+        // misaddressed. Pin this agent's own entry per-spawn (the launch builders
+        // emit `-c mcp_servers.threadline.*`). Only when threadline is configured;
+        // ignored by non-codex launches. See CODEX-MULTIAGENT-THREADLINE-SPEC.
+        const codexThreadlineMcp = config.threadline
+            ? resolveThreadlineMcpEntry(config.sessions.projectDir, config.stateDir, config.projectName)
+            : undefined;
+        const sessionManager = new SessionManager(codexThreadlineMcp ? { ...config.sessions, codexThreadlineMcp } : config.sessions, state);
         // Input Guard is constructed later (after sharedIntelligence is available)
         // so the topic coherence reviewer can route through the IntelligenceProvider
         // abstraction instead of calling Anthropic directly.
@@ -5970,6 +5981,12 @@ export async function startServer(options) {
                     // §4.5: honor SpawnRequestManager's provenance tag so drain-spawned
                     // sessions are distinguishable from inline-spawned ones in logs/stream.
                     triggeredBy: opts?.triggeredBy ?? 'spawn-request',
+                    // This is the Threadline inbound-reply spawn: the worker must call
+                    // the threadline_send MCP tool to reply, which a codex worker can only
+                    // do under full bypass (codex cancels MCP calls in any sandbox). Jobs
+                    // do NOT set this and stay sandboxed. Bounded: Threadline only accepts
+                    // messages from trusted agents.
+                    codexAllowMcpTools: true,
                 });
                 return session.id;
             },
@@ -6160,6 +6177,9 @@ export async function startServer(options) {
                     minIqsBand: pipeConfig?.minIqsBand ?? 70,
                     framework: pipeFramework,
                     binaryPath: pipeBinaryPath,
+                    // Same per-agent codex threadline MCP override as SessionManager, so a
+                    // codex pipe-reply worker uses THIS agent's threadline MCP.
+                    ...(codexThreadlineMcp ? { codexThreadlineMcp } : {}),
                 });
                 console.log(pc.dim(`  Pipe sessions: enabled (model: ${pipeConfig?.model ?? 'sonnet'}, max: ${pipeConfig?.maxConcurrent ?? 5})`));
             }
@@ -6579,6 +6599,15 @@ export async function startServer(options) {
         else {
             console.log(pc.yellow('  Discovery evaluator: inactive (no IntelligenceProvider)'));
         }
+        // Independent autonomous-completion judge (mirrors /goal). Reuses the
+        // framework-aware sharedIntelligence; falls back to the self-declared promise
+        // when absent (the stop-hook handles that).
+        let completionEvaluator;
+        if (sharedIntelligence) {
+            const { CompletionEvaluator } = await import('../core/CompletionEvaluator.js');
+            completionEvaluator = new CompletionEvaluator({ intelligence: sharedIntelligence });
+            console.log(pc.green('  Completion evaluator: active (independent /goal-style judge)'));
+        }
         // Register feature-discovery probe for self-knowledge tree (Phase 4: Agent Integration)
         if (selfKnowledgeTree && featureRegistry) {
             selfKnowledgeTree.probes.register('feature-discovery', async () => {
@@ -6911,7 +6940,7 @@ export async function startServer(options) {
                 },
             });
         }
-        const server = new AgentServer({ config, sessionManager, state, scheduler, telegram, relationships, feedback, feedbackAnomalyDetector, dispatches, updateChecker, autoUpdater, autoDispatcher, quotaTracker, quotaManager, publisher, viewer, tunnel, evolution, watchdog, topicMemory, triageNurse, projectMapper, coherenceGate: scopeVerifier, contextHierarchy, canonicalState, operationGate, sentinel, adaptiveTrust, memoryMonitor, orphanReaper, coherenceMonitor, commitmentTracker, semanticMemory, activitySentinel, rateLimitSentinel, messageRouter, summarySentinel, spawnManager, systemReviewer, capabilityMapper, selfKnowledgeTree, coverageAuditor, topicResumeMap: _topicResumeMap ?? undefined, sessionRefresh: _sessionRefresh ?? undefined, autonomyManager, trustElevationTracker, autonomousEvolution, coordinator: coordinator.enabled ? coordinator : undefined, localSigningKeyPem, whatsapp: whatsappAdapter, slack: slackAdapter, imessage: imessageAdapter, whatsappBusinessBackend, messageBridge, hookEventReceiver, worktreeMonitor, subagentTracker, instructionsVerifier, handshakeManager: threadlineHandshake, threadlineRouter, threadResumeMap, topicLinkageHandler: topicLinkageHandler ?? undefined, threadlineRelayClient, threadlineReplyWaiters, listenerManager: listenerManager ?? undefined, responseReviewGate, messagingToneGate, outboundDedupGate, telemetryHeartbeat, pasteManager, featureRegistry, discoveryEvaluator, unifiedTrust, liveConfig, sharedStateLedger, ledgerSessionRegistry, worktreeManager, oidcEnrolledRepos: parallelDevConfig?.oidcEnrolledRepos, initiativeTracker, projectRoundRunner, projectDriftChecker, machineHeartbeat, proxyCoordinator, topicIntentStore, telegramBridgeConfig, telegramBridge: telegramBridge ?? undefined, threadlineObservability, workingMemory, taskFlowRegistry, threadlineFlowBridge });
+        const server = new AgentServer({ config, sessionManager, state, scheduler, telegram, relationships, feedback, feedbackAnomalyDetector, dispatches, updateChecker, autoUpdater, autoDispatcher, quotaTracker, quotaManager, publisher, viewer, tunnel, evolution, watchdog, topicMemory, triageNurse, projectMapper, coherenceGate: scopeVerifier, contextHierarchy, canonicalState, operationGate, sentinel, adaptiveTrust, memoryMonitor, orphanReaper, coherenceMonitor, commitmentTracker, semanticMemory, activitySentinel, rateLimitSentinel, messageRouter, summarySentinel, spawnManager, systemReviewer, capabilityMapper, selfKnowledgeTree, coverageAuditor, topicResumeMap: _topicResumeMap ?? undefined, sessionRefresh: _sessionRefresh ?? undefined, autonomyManager, trustElevationTracker, autonomousEvolution, coordinator: coordinator.enabled ? coordinator : undefined, localSigningKeyPem, whatsapp: whatsappAdapter, slack: slackAdapter, imessage: imessageAdapter, whatsappBusinessBackend, messageBridge, hookEventReceiver, worktreeMonitor, subagentTracker, instructionsVerifier, handshakeManager: threadlineHandshake, threadlineRouter, threadResumeMap, topicLinkageHandler: topicLinkageHandler ?? undefined, threadlineRelayClient, threadlineReplyWaiters, listenerManager: listenerManager ?? undefined, responseReviewGate, messagingToneGate, outboundDedupGate, telemetryHeartbeat, pasteManager, featureRegistry, discoveryEvaluator, completionEvaluator, unifiedTrust, liveConfig, sharedStateLedger, ledgerSessionRegistry, worktreeManager, oidcEnrolledRepos: parallelDevConfig?.oidcEnrolledRepos, initiativeTracker, projectRoundRunner, projectDriftChecker, machineHeartbeat, proxyCoordinator, topicIntentStore, telegramBridgeConfig, telegramBridge: telegramBridge ?? undefined, threadlineObservability, workingMemory, taskFlowRegistry, threadlineFlowBridge });
         // Boot-recovery (tunnel-failure-resilience spec Part 6): if the agent
         // died mid-relay-episode, the persisted tunnel.json carries
         // rotationPending=true. Rotate the dashboard PIN + authToken BEFORE