@amistio/cli 0.1.31 → 0.1.32

package/README.md

@@ -17,6 +17,8 @@ Runner Update installs the official `@amistio/cli` package and then refreshes th
 
 Current runners advertise the work kinds they can claim. Older runners that do not send this capability can continue legacy brain generation, implementation, and plan revision work, but they will skip source-aware assistant, impact-preview, semantic brain consolidation, project-context refresh, issue-diagnosis, app-evaluation, security-posture, Test-quality, implementation-Test-gate, and implementation-verification work until updated.
 
+Tool session reuse is bounded. One-shot tool sessions close after successful completion, failed runs are blocked, active sessions are treated as already in use, and reusable sessions idle for more than six hours are closed before the next claim selects context. This keeps follow-up work from inheriting stale local AI context while preserving recent reusable sessions for related work.
+
 Repository brain auto-sync is disabled until the repository link option is enabled in the app. After pairing, run `amistio sync watch` from the paired checkout to push recognized external brain Markdown/MDX files and explicit HTML artifacts under `docs/html/<area>/`, including local ADRs, plans, prompts, workflows, memory, context, architecture, and feature docs, to the app for review. Markdown is the default generation format; HTML appears only when a runner or user explicitly generated an HTML artifact. `amistio run --watch` also runs the same cycle between work polls when the option is enabled. The CLI skips templates, unsupported paths, oversized files, unchanged managed docs, and conflicts instead of silently overwriting web state.
 
 Repository autopilot is disabled until the repository link option is enabled in the app. When enabled, Amistio can attach an audited low-risk autopilot authorization to eligible runner work, including generated brain approval, impact preview, issue diagnosis, security posture scan, app evaluation cleanup, low-risk implementation handoff, requeue, and implementation verification. The Runner panel shows and updates safe work scopes, allowed candidate types, max risk, optional runner binding, daily/concurrent/failure budgets, expiry/review/cooldown windows, and pause state. The CLI shows authorization id, candidate id/type, outcome, policy version, and work kind in `amistio work list`, claim logs, runner prompts, and milestone activity. Autopilot does not widen local runner permissions: pairing, supported work kinds, runner identity, Git worktree isolation, redaction, local-tool permission controls, and unsafe/review-required/blocked/paused/budget stops still apply.

package/dist/index.js

@@ -4740,8 +4740,32 @@ function runProcess(command, args, timeoutMs) {
   });
 }
 
+// src/tool-session-lifecycle.ts
+var TOOL_SESSION_MAX_IDLE_MS = 6 * 60 * 60 * 1e3;
+function completedToolSessionStatus(session) {
+  return session.resumabilityScope === "none" ? "closed" : "open";
+}
+function completedToolSessionClosedReason(session) {
+  if (session.resumabilityScope !== "none") {
+    return void 0;
+  }
+  return "Completed one-shot tool run; this session is not reusable.";
+}
+function staleToolSessionClosedReason(session, now = /* @__PURE__ */ new Date()) {
+  if (session.status !== "open") {
+    return void 0;
+  }
+  const lastActivityMs = Date.parse(session.lastActivityAt);
+  if (!Number.isFinite(lastActivityMs)) {
+    return "Session has an invalid last activity timestamp; closing to prevent stale context reuse.";
+  }
+  if (lastActivityMs + TOOL_SESSION_MAX_IDLE_MS >= now.getTime()) {
+    return void 0;
+  }
+  return "Session idle window expired; closing to prevent stale context reuse.";
+}
+
 // src/session-policy.ts
-var maxIdleMs = 24 * 60 * 60 * 1e3;
 var maxTotalMs = 7 * 24 * 60 * 60 * 1e3;
 var maxMessageCount = 80;
 var maxEstimatedTokens = 12e4;
@@ -4825,10 +4849,10 @@ function sessionIneligibleReason(session, input, now) {
   if (input.workItem.executionWorktreeKey && session.executionWorktreeKey !== input.workItem.executionWorktreeKey) {
     return "worktree scope mismatch";
   }
-  if (session.status === "closed" || session.status === "archived" || session.status === "blocked" || session.status === "unavailable") {
+  if (session.status === "closed" || session.status === "archived" || session.status === "blocked" || session.status === "unavailable" || session.status === "active") {
     return `session is ${session.status}`;
   }
-  if (Date.parse(session.lastActivityAt) + maxIdleMs < now.getTime()) {
+  if (Date.parse(session.lastActivityAt) + TOOL_SESSION_MAX_IDLE_MS < now.getTime()) {
     return "session is idle past the reuse window";
   }
   if (Date.parse(session.createdAt) + maxTotalMs < now.getTime()) {
@@ -6035,6 +6059,7 @@ function createAppEvaluationScanPrompt(workItem, context) {
     "- Treat active proposed, approved, ready, or in-progress plans/prompts with accepted controlling ADRs, features, or work artifacts as active backlog. Do not classify them as cleanup material solely because they are older, unexecuted, or not yet fully implemented; use keepActive when they still describe valid pending or approved work.",
     "- Treat intentionally in-progress feature tracks as still-active work when their controlling plan/feature has unchecked requirements or explicit follow-up gaps. For example, a completed first implementation prompt does not make the broader feature stale if PLAN/FEAT evidence says remaining lifecycle work is still open; return proposedLifecycleAction keepActive with evidence instead of cleanup.",
     "- Treat prompt frontmatter status Ready as an active execution backlog state by default, not as stale review debt. Only flag a Ready prompt for metadata correction when its controlling plan, feature, prompt index, or verification evidence unambiguously proves the prompt has already completed or been superseded.",
+    "- Treat implemented umbrella plans that explicitly label unchecked checklist items as deferred follow-ups, future candidates, roadmap backlog, or split-out hardening phases as valid deferred backlog. Do not mark the umbrella incomplete or stale solely because those deferred items remain unchecked; use keepActive or no cleanup, and recommend a fresh focused plan only when a concrete deferred slice has current evidence and approval.",
     "- When lifecycle metadata disagrees across indexes, frontmatter, feature specs, ADRs, and implementation evidence, cite the conflict and propose a metadata correction or verification step instead of archival/removal.",
     "- Check missing memory or workflow updates when repeated lessons or operational rules are visible.",
     "- Check release readiness, UX, accessibility, performance, reliability, and security-posture follow-through at a summary level.",
@@ -11026,6 +11051,8 @@ async function prepareToolSession({
   workItem
 }) {
   const { toolSessions } = await apiClient.listToolSessions(projectId);
+  const now = /* @__PURE__ */ new Date();
+  await closeStaleToolSessionsBestEffort(apiClient, projectId, toolSessions, now);
   const selection = selectToolSession({
     policy: sessionPolicy,
     workItem,
@@ -11034,7 +11061,8 @@ async function prepareToolSession({
     runnerId,
     repositoryLinkId,
     machineId,
-    supportsSessionReuse
+    supportsSessionReuse,
+    now
   });
   if (selection.decision === "skipped") {
     return selection;
@@ -11076,6 +11104,18 @@ async function prepareToolSession({
   });
   return { ...selection, toolSession };
 }
+async function closeStaleToolSessionsBestEffort(apiClient, projectId, toolSessions, now) {
+  const staleSessions = toolSessions.map((session) => ({ session, reason: staleToolSessionClosedReason(session, now) })).filter((item) => Boolean(item.reason));
+  if (!staleSessions.length) {
+    return;
+  }
+  const settlements = await Promise.allSettled(staleSessions.map(({ session, reason }) => apiClient.updateToolSession(projectId, session.toolSessionId, {
+    status: "closed",
+    closedReason: reason,
+    summary: session.summary ?? reason
+  })));
+  logRejectedSettlements("close stale tool sessions", settlements);
+}
 async function finalizeToolSession({
   apiClient,
   costUsd,
@@ -11093,8 +11133,10 @@ async function finalizeToolSession({
     return void 0;
   }
   const summary = summarizeToolOutput(stdout) ?? session.summary;
+  const nextStatus = status === "completed" ? completedToolSessionStatus(session) : "blocked";
+  const closedReason = status === "completed" ? completedToolSessionClosedReason(session) : "Last run failed or returned a non-zero exit code.";
   const { toolSession } = await apiClient.updateToolSession(projectId, session.toolSessionId, {
-    status: status === "completed" ? "open" : "blocked",
+    status: nextStatus,
     runnerId,
     lastWorkItemId: workItemId,
     messageCount: (session.messageCount ?? 0) + (messageCount ?? 1),
@@ -11102,7 +11144,7 @@ async function finalizeToolSession({
     ...tokensIn !== void 0 ? { estimatedInputTokens: (session.estimatedInputTokens ?? 0) + tokensIn } : {},
     ...tokensOut !== void 0 ? { estimatedOutputTokens: (session.estimatedOutputTokens ?? 0) + tokensOut } : {},
     ...costUsd !== void 0 ? { costUsd: (session.costUsd ?? 0) + costUsd } : {},
-    ...status === "failed" ? { closedReason: "Last run failed or returned a non-zero exit code." } : {}
+    ...closedReason ? { closedReason } : {}
   });
   return toolSession;
 }