sortie-dogs 0.9.7 → 0.9.8

package/README.md

@@ -1,13 +1,13 @@
 # Sortie-dogs
 
-**Give OpenCode a task; get a bounded, validated implementation loop instead of an open-ended agent run.**
+**Add a bounded, cost-aware execution loop to OpenCode without taking OpenCode over.**
 
-Sortie-dogs is an execution harness for OpenCode, applying harness-engineering
-principles through mechanical write boundaries, validation, review, and bounded
-orchestration.
+Sortie-dogs is an opt-in overlay, not a replacement. It stays passive until you
+invoke `/sortie` or select `dog-coordinator`. Standard OpenCode agents and
+unrelated sessions remain available and unchanged.
 
-> **Project status: Beta.** Stable in regular use. As a pre-1.0 release,
-> configuration and runtime assets may still change between releases.
+> **Project status: Beta.** v0.9.x is under active stabilization. Runtime
+> behavior, configuration, and runtime assets may still change before 1.0.
 
 [![npm](https://img.shields.io/npm/v/sortie-dogs)](https://www.npmjs.com/package/sortie-dogs)
 [![license](https://img.shields.io/npm/l/sortie-dogs)](LICENSE)
@@ -15,47 +15,63 @@ orchestration.
 
 ![Sortie-dogs coordinating a bounded implementation workflow](https://raw.githubusercontent.com/zufall-upon/Sortie-dogs/main/docs/assets/sortie-workflow.gif)
 
-Sortie-dogs is an opt-in OpenCode orchestration plugin. It turns a task into a
-scoped plan, optional evidence gathering, one bounded implementation unit, canonical
-validation, and evidence-backed completion—while preserving standard OpenCode
-agents and settings.
+Sortie-dogs turns selected work into a scoped plan, optional evidence gathering,
+bounded implementation, canonical validation, and evidence-backed completion.
 
 Requirements: Node.js 22.6 or newer, npm, and OpenCode.
 
 Guides: [日本語](docs/guide-ja.md) · [简体中文](docs/guide-zh-CN.md) · [CLI testing](docs/cli-testing.md)
 
-Release: [v0.9.7](https://github.com/zufall-upon/Sortie-dogs/releases/tag/v0.9.7)
-
-## Quick start
-
-Install the public npm package in the project and generate the project-local
-OpenCode runtime files:
+Release: [v0.9.8](https://github.com/zufall-upon/Sortie-dogs/releases/tag/v0.9.8)
+
+## Why Sortie-dogs?
+
+### Invisible until invited
+
+Use normal OpenCode normally. Sortie activates only for `/sortie` or
+`dog-coordinator`; it does not disable or replace OpenCode's standard agents.
+
+### Spend strong models only where they matter
+
+Lower-cost models handle bounded retrieval and parallel volume work. Stronger
+models are reserved for implementation, escalation, and independent review.
+
+### Return with proof
+
+Writes stay scoped, and completion requires validation evidence. Every completed
+run can return a concise Speed / Cost / Proof debrief.
+
+## Designed to coexist with OpenCode
+
+Sortie-dogs adds a workflow to your existing setup rather than replacing it.
+
+- It does not disable OpenCode's native agents or replace standard roles such as
+  `build`, `plan`, `explore`, or `general`.
+- Ordinary sessions are not automatically converted into Sortie workflows.
+- Project-local initialization is the recommended setup and does not change
+  user settings.
+- Global runtime availability and cross-project reflection require separate,
+  explicit opt-in. Reflection is disabled by default.
+- Unknown or user-owned runtime files are preserved rather than overwritten.
+- Manual removal targets only known Sortie-owned runtime assets.
+
+Use OpenCode normally. Invoke the pack only when you want it.
+
+## Quick start
+
+Install the public npm package in the project and generate the project-local
+OpenCode runtime files:
 
 ```sh
 npm install --save-dev sortie-dogs
-npx sortie-dogs init .
-```
-
-Alternatively, install the CLI globally and initialize OpenCode's global
-configuration:
-
-```sh
-npm install --global sortie-dogs
-sortie-dogs init --global
-```
-
-This installs the canonical runtime assets in OpenCode's global configuration,
-so `dog-coordinator` can be selected from other projects without project-local
-initialization. Global initialization and project-local initialization are
-separate: `sortie-dogs init .` still writes runtime files only into that
-project. Project-local configuration and the plugin bridge below remain
-available when a project needs its own settings or dependency.
-
-Installing the runtime assets does not load the plugin, and without the plugin
-every role runs on whichever model the caller happened to use. Add the package
-to the `plugin` array of the OpenCode configuration the agents run under —
-`~/.config/opencode/opencode.json` for the global assets, or the project's
-`.opencode/opencode.json`:
+npx sortie-dogs init .
+```
+
+This is the recommended setup. Runtime assets stay project-local.
+
+Installing the runtime assets does not load the plugin, and without the plugin
+every role runs on whichever model the caller happened to use. Add the package
+to the `plugin` array in the project's `.opencode/opencode.json`:
 
 ```json
 {
@@ -218,26 +234,7 @@ global file for durable global settings.
   unit, with a maximum of three records per run; routine bugs and external
   failures are never journaled.
 
-## Why Sortie-dogs
-
-- **Focused when invited, invisible otherwise.** Activate it with `/sortie` or
-  select `dog-coordinator`; ordinary OpenCode sessions remain unchanged.
-- **Evidence gathering without arbitrary dispatch ceilings.** A bounded scout resolves
-  each concrete manifest, validation, or ownership gap whenever it appears; unchanged
-  requests are not repeated.
-- **Writes stay inside the assignment.** Exact source or operation manifests
-  gate edits and handoffs.
-- **Autonomous implementation routing.** An accepted scope with at least two safe independent
-  units defaults to Luna fabric without opt-in. An explicit serial/no-parallel request wins;
-  other work uses sequential workers. The separate parallel contract remains explicit.
-- **Runtime overlap protection.** Active equal or ancestor write scopes are rejected
-  before mutation, and full validation waits until every parallel unit joins.
-- **Evidence before completion.** Canonical validation, risk-based review, and
-  terminal evidence gate coordinator-owned completion and commits.
-- **Long work can recover.** Restart recovery and bounded compaction continue
-  from retained handoff context rather than silently starting over.
-
-## Example run
+## Example run
 
 An illustrative low-risk run stays bounded and reports its gates:
 
@@ -400,7 +397,30 @@ coordinator. `dog-reviewer` independently checks high-risk candidates after
 canonical validation. Neither role implements, stages, commits, or acts as a
 user-facing worker.
 
-## Updates and migration
+## Optional global availability
+
+If you intentionally want the Sortie roles available across projects, install
+the CLI and runtime assets globally:
+
+```sh
+npm install --global sortie-dogs
+sortie-dogs init --global
+```
+
+This writes canonical Sortie runtime assets to OpenCode's global configuration;
+it does not make project-local initialization global. Load the plugin from the
+global `~/.config/opencode/opencode.json` when using these assets:
+
+```json
+{
+  "plugin": ["sortie-dogs"]
+}
+```
+
+Project-local configuration and plugin loading remain available when a project
+needs its own settings or dependency.
+
+## Updates and migration
 
 After replacing the dependency with a newer release asset, run:
 

package/dist/asset-version.d.ts

@@ -2,5 +2,5 @@
  * Version of the installable runtime assets. Kept in its own module so the plugin can compare an
  * installed project marker without importing every asset body.
  */
-export declare const RUNTIME_ASSET_VERSION = "0.3.81-mission-debrief-v1";
+export declare const RUNTIME_ASSET_VERSION = "0.3.84-dispatch-recovery-v1";
 export type RuntimeAssetVersion = typeof RUNTIME_ASSET_VERSION;

package/dist/asset-version.js

@@ -2,4 +2,4 @@
  * Version of the installable runtime assets. Kept in its own module so the plugin can compare an
  * installed project marker without importing every asset body.
  */
-export const RUNTIME_ASSET_VERSION = "0.3.81-mission-debrief-v1";
+export const RUNTIME_ASSET_VERSION = "0.3.84-dispatch-recovery-v1";

package/dist/core/goal-bound.d.ts

@@ -243,7 +243,7 @@ export interface GoalFlightState {
 }
 export declare class GoalBoundError extends Error {
     readonly code: "invalid" | "transition" | "budget" | "ticket" | "evidence";
-    constructor(code: "invalid" | "transition" | "budget" | "ticket" | "evidence", message: string);
+    constructor(code: GoalBoundError["code"], message: string);
 }
 export declare function goalFingerprint(value: unknown): string;
 export declare function selectGoalDelivery(input: {

package/dist/core/goal-declaration-format.d.ts

@@ -2,3 +2,5 @@
 export declare const GOAL_DELIVERY_INTENTS: readonly ["design", "registration", "implementation", "repair", "controlled-change"];
 export declare const GOAL_DELIVERY_MODES: readonly ["planning-only", "mvp-first", "repair-first", "controlled-change"];
 export declare const GOAL_DECLARATION_FORMAT: string;
+/** Normalize a shared declaration without guessing acceptance, coverage, delivery, or validation. */
+export declare function expandGoalDeclaration(value: unknown): string;

package/dist/core/goal-declaration-format.js

@@ -1,12 +1,16 @@
+import { goalFingerprint } from "./goal-bound.js";
 /** Shared planner guidance and admission enums; no inferred acceptance or dispatch authority. */
 export const GOAL_DELIVERY_INTENTS = ["design", "registration", "implementation", "repair", "controlled-change"];
 export const GOAL_DELIVERY_MODES = ["planning-only", "mvp-first", "repair-first", "controlled-change"];
-export const GOAL_DECLARATION_FORMAT = `Goal declaration belongs in the Task prompt, not in the handoff JSON or operation manifest.
+export const GOAL_DECLARATION_FORMAT = `Declare a goal once, then reference it with goal_declaration_path in the Task prompt.
+The referenced JSON may contain shared defaults and a criteria array; the host expands them privately.
+An inline ext["sortie-dogs/goal-declaration"] in the registered handoff is also supported.
+Existing accepted goals need no repeated declaration; budget-only revisions retain their criteria.
+Legacy flat goal_* fields remain supported. No need to rewrite unrelated files or repeat common fields for each criterion.
 delivery_intent must be exactly one of: ${GOAL_DELIVERY_INTENTS.join(" | ")}. Never use prose or a boolean.
 delivery_mode is a separate optional enum: ${GOAL_DELIVERY_MODES.join(" | ")}.
 Use flat key: value lines. Each criterion begins with its own goal_criterion_id: line.
-Repeat the entire flat criterion block for multiple criteria. Do not use goal_criteria, goal_acceptance.criteria,
-YAML object/list wrappers, inline { goal_criterion_id: ... }, or JSON objects for these blocks.
+For legacy flat input each criterion has a goal_criterion_id line. Prefer shared JSON defaults for multiple criteria.
 Replace every placeholder with the accepted task's actual value; preserve its acceptance and validation:
 goal_acceptance_fingerprint: <sha256: followed by exactly 64 lowercase hexadecimal characters>
 delivery_intent: implementation
@@ -31,3 +35,48 @@ goal_validation_command: <exact operation manifest validation command>
 The example's implementation/mvp-first selections apply only to an implementation without its usable path;
 select other enum values only from the accepted request. A declaration rejection launches no child.
 Repair the Task prompt's named fields, not unrelated files; retry only with the corrected declaration.`;
+/** Normalize a shared declaration without guessing acceptance, coverage, delivery, or validation. */
+export function expandGoalDeclaration(value) {
+    if (value === null || typeof value !== "object" || Array.isArray(value))
+        throw new Error("goal declaration must be an object");
+    const object = value;
+    if (!Array.isArray(object.criteria) || object.criteria.length === 0)
+        throw new Error("goal declaration requires criteria");
+    const defaults = object.defaults !== null && typeof object.defaults === "object" && !Array.isArray(object.defaults)
+        ? object.defaults : {};
+    const fields = ["criterion_id", "target", "entrypoint", "workload", "oracle_coverage", "build_boundary", "source",
+        "candidate", "source_binding", "candidate_binding", "fixture", "proof_scope", "expected_outcome", "validation_command"];
+    const scalar = (value) => {
+        if (Array.isArray(value))
+            return JSON.stringify(value);
+        if (typeof value === "string" && !/[\r\n]/u.test(value))
+            return value;
+        if (typeof value === "boolean" || typeof value === "number")
+            return String(value);
+        throw new Error("goal declaration field must be a single-line scalar or array");
+    };
+    const normalizedFields = (values) => Object.fromEntries(fields.flatMap(field => {
+        const found = values[`goal_${field}`] ?? values[field];
+        return found === undefined ? [] : [[`goal_${field}`, found]];
+    }));
+    const criteria = object.criteria.map(raw => {
+        if (raw === null || typeof raw !== "object" || Array.isArray(raw))
+            throw new Error("goal criterion must be an object");
+        return { ...normalizedFields(defaults), ...normalizedFields(raw) };
+    });
+    const identity = { criteria, delivery_intent: object.delivery_intent, delivery_mode: object.delivery_mode,
+        usable_path_established: object.usable_path_established, controlled_change: object.controlled_change };
+    const lines = [`goal_acceptance_fingerprint: ${scalar(object.goal_acceptance_fingerprint ?? goalFingerprint(identity))}`];
+    for (const field of ["delivery_intent", "delivery_mode", "usable_path_established", "controlled_change",
+        "goal_budget_units", "goal_budget_time_ms", "goal_budget_cost_usd"]) {
+        if (object[field] !== undefined)
+            lines.push(`${field}: ${scalar(object[field])}`);
+    }
+    for (const criterion of criteria) {
+        criterion.goal_criterion_id ??= `criterion-${goalFingerprint(criterion).slice(7, 31)}`;
+        for (const field of fields)
+            if (criterion[`goal_${field}`] !== undefined)
+                lines.push(`goal_${field}: ${scalar(criterion[`goal_${field}`])}`);
+    }
+    return lines.join("\n");
+}

package/dist/plugin/index.d.ts

@@ -65,6 +65,8 @@ export interface OpenCodeHooks {
     /** Continuation observes the coordinator's completed final text to honour its fallback markers. */
     "experimental.text.complete"?: (input: {
         sessionID: string;
+        messageID?: string;
+        partID?: string;
     }, output: {
         text: string;
     }) => Promise<void>;

package/dist/plugin/index.js

@@ -10,7 +10,7 @@ import { createHash, randomUUID } from "node:crypto";
 import { lstat, mkdir, open, readFile, readdir, realpath, rm, stat, writeFile } from "node:fs/promises";
 import { basename, dirname, isAbsolute, join, relative, resolve } from "node:path";
 import { RUNTIME_ASSET_VERSION } from "../asset-version.js";
-import { GOAL_DECLARATION_FORMAT, GOAL_DELIVERY_INTENTS, GOAL_DELIVERY_MODES } from "../core/goal-declaration-format.js";
+import { GOAL_DECLARATION_FORMAT, GOAL_DELIVERY_INTENTS, GOAL_DELIVERY_MODES, expandGoalDeclaration } from "../core/goal-declaration-format.js";
 import { ACCEPTANCE_CONTINUITY_AUTHORITY, ACCEPTANCE_CONTINUITY_EXTENSION, ACCEPTANCE_CONTINUITY_SCHEMA_VERSION, acceptanceContinuityFingerprint, inspectAcceptanceContinuity, normalizeAcceptanceCriteria, } from "../core/acceptance-continuity.js";
 import { resolveGlobalConfigRoot } from "../core/initialize.js";
 import { admitLunaFabric } from "../core/luna-fabric-contract.js";
@@ -1699,6 +1699,17 @@ export const SortieDogsPlugin = async (input, options) => {
             rootAcceptanceContinuity.delete(sessionID);
         return { outcome, goal, receipt, delivery, records: snapshot?.records };
     }
+    const ROOT_INTERRUPTION = /(^|\n)TRUE_INTERRUPTION\s*:\s*(?:user|internal)\s*:/iu;
+    async function preserveActiveGoalContinuation(sessionID, text, messageID) {
+        if (!isCoordinatorSession(sessionID) || terminalRunOutcome(text) !== "INTERRUPTED" ||
+            ROOT_INTERRUPTION.test(text) || hasCoordinatorInterruption(sessionID, messageID)) {
+            return text;
+        }
+        const goal = await currentGoal(sessionID).catch(() => undefined);
+        if (goal === undefined || goal.goal_id === null || goal.receipt !== null)
+            return text;
+        return replaceTerminalStatus(text, "status: IN_PROGRESS — local/process/step continuation remains active in the same session");
+    }
     function goalDeclarationContract(prompt) {
         const lines = prompt.split(/\r?\n/u);
         const criterionStarts = lines.flatMap((line, index) => /^\s*goal_criterion_id\s*:/u.test(line) ? [index] : []);
@@ -1822,15 +1833,52 @@ export const SortieDogsPlugin = async (input, options) => {
         });
         return { declaration: { fingerprint, delivery, contract: contract.contract }, defects };
     }
+    function resolveGoalDeclaration(state, prompt) {
+        const keys = [...handoffEntries(prompt).keys()].filter(key => /^(?:goal_|delivery_intent$|delivery_mode$|usable_path_established$|controlled_change$)/u.test(key));
+        const budgetOnly = keys.every(key => ["goal_budget_units", "goal_budget_time_ms", "goal_budget_cost_usd",
+            "goal_acceptance_fingerprint"].includes(key));
+        const fingerprint = handoffValue(handoffEntries(prompt), ["goal_acceptance_fingerprint"]);
+        if (budgetOnly && state.acceptance_contract !== null && state.acceptance_fingerprint !== null && state.delivery !== null &&
+            (fingerprint === undefined || fingerprint === state.acceptance_fingerprint)) {
+            return { declaration: { fingerprint: state.acceptance_fingerprint, delivery: state.delivery,
+                    contract: state.acceptance_contract }, defects: [], inherited: true };
+        }
+        return { ...validateGoalDeclaration(prompt), inherited: false };
+    }
+    async function resolveGoalPrompt(prompt) {
+        const entries = handoffEntries(prompt);
+        if (entries.has("goal_criterion_id"))
+            return prompt;
+        const reference = handoffValue(entries, ["goal_declaration_path"]);
+        let definition;
+        if (reference !== undefined)
+            definition = await readJson(resolve(input.directory, reference), INPUT_LIMITS.handoff);
+        else {
+            const handoffPath = handoffValue(entries, ["handoff_path", "handoffpath"]);
+            if (handoffPath !== undefined) {
+                const handoff = await readJson(resolve(input.directory, handoffPath), INPUT_LIMITS.handoff);
+                if (isRecord(handoff) && isRecord(handoff.ext))
+                    definition = handoff.ext["sortie-dogs/goal-declaration"];
+            }
+        }
+        if (definition === undefined)
+            return prompt;
+        const expanded = expandGoalDeclaration(definition).split("\n").filter(line => {
+            const field = line.slice(0, line.indexOf(":"));
+            return !entries.has(field);
+        });
+        return `${prompt}\n${expanded.join("\n")}`;
+    }
     async function bindGoalDeclaration(sessionID, prompt) {
+        prompt = await resolveGoalPrompt(prompt);
         const ledger = await goalLedger(sessionID);
         let state = (await ledger.readGoal()).state;
         if (state.goal_id === null || state.origin_user_message_id === null)
             return undefined;
         const entries = handoffEntries(prompt);
         const typedDeclarationPresent = prompt.split(/\r?\n/u).some((line) => /^\s*(?:goal_[a-z0-9_]+|delivery_intent|delivery_mode|usable_path_established|controlled_change)\s*:/iu.test(line));
-        // Accepted goals may retain their existing declaration on later units, but the first worker
-        // handoff and every handoff containing declaration fields must be complete.
+        // A budget-only amendment retains the accepted goal contract, just like a later unit with
+        // no declaration fields. Explicit acceptance changes still use the existing full declaration.
         if (!typedDeclarationPresent) {
             if (state.acceptance_contract === null && goalDeclarationAuthority.get(sessionID) === state.latest_user_message_id) {
                 throw new HandoffDeniedError("contract-invalid", "<goal-declaration>", {
@@ -1841,7 +1889,7 @@ export const SortieDogsPlugin = async (input, options) => {
                 goalDeclarationAuthority.delete(sessionID);
             return state;
         }
-        const validated = validateGoalDeclaration(prompt);
+        const validated = resolveGoalDeclaration(state, prompt);
         if (validated.declaration === undefined) {
             throw new HandoffDeniedError("contract-invalid", "<goal-declaration>", { defects: validated.defects });
         }
@@ -1883,6 +1931,7 @@ export const SortieDogsPlugin = async (input, options) => {
     async function reserveGoalDispatch(sessionID, callID, prompt) {
         if (goalReservations.has(callID))
             return;
+        await recoverCompletedGoalReservations(sessionID);
         const ledger = await goalLedger(sessionID);
         let state = await bindGoalDeclaration(sessionID, prompt) ?? (await ledger.readGoal()).state;
         if (state.goal_id === null)
@@ -2064,6 +2113,26 @@ export const SortieDogsPlugin = async (input, options) => {
     }), { issueTicket: issueGoalTicket });
     const completedCoordinatorMessages = new Set();
     const completedCoordinatorParts = new Set();
+    const interruptedCoordinatorMessages = new Map();
+    function rememberCoordinatorInterruption(sessionID, info) {
+        const error = isRecord(info.error) ? info.error : undefined;
+        if (info.role !== "assistant" || info.agent !== COORDINATOR_AGENT || typeof info.id !== "string" ||
+            error?.name !== "MessageAbortedError")
+            return;
+        const messages = interruptedCoordinatorMessages.get(sessionID) ?? new Set();
+        messages.add(info.id);
+        interruptedCoordinatorMessages.delete(sessionID);
+        interruptedCoordinatorMessages.set(sessionID, messages);
+        while (messages.size > ACTIVE_SESSION_CACHE.maximum)
+            messages.delete(messages.values().next().value);
+        while (interruptedCoordinatorMessages.size > ACTIVE_SESSION_CACHE.maximum) {
+            interruptedCoordinatorMessages.delete(interruptedCoordinatorMessages.keys().next().value);
+        }
+    }
+    function hasCoordinatorInterruption(sessionID, messageID) {
+        const messages = interruptedCoordinatorMessages.get(sessionID);
+        return messages !== undefined && (messageID === undefined || messages.has(messageID));
+    }
     function freshSessionFallback(reason, action) {
         return { status: "user-action-required", reason, action, retry_same_session: false };
     }
@@ -4734,6 +4803,9 @@ export const SortieDogsPlugin = async (input, options) => {
         return true;
     }
     function expireSession(sessionID) {
+        for (const key of inspectionOperations.keys())
+            if (key.startsWith(`${sessionID}\u0000`))
+                inspectionOperations.delete(key);
         activeSessions.delete(sessionID);
         sessionOperationMetrics.delete(sessionID);
         abandonSessionLease(sessionID);
@@ -4755,6 +4827,9 @@ export const SortieDogsPlugin = async (input, options) => {
         clearSessionLinks(sessionID);
     }
     function evictSession(sessionID) {
+        for (const key of inspectionOperations.keys())
+            if (key.startsWith(`${sessionID}\u0000`))
+                inspectionOperations.delete(key);
         goalValidationDefects.delete(sessionID);
         activeSessions.delete(sessionID);
         sessionOperationMetrics.delete(sessionID);
@@ -4771,6 +4846,7 @@ export const SortieDogsPlugin = async (input, options) => {
         assetVersionPins.delete(sessionID);
         coordinatorPrompts.delete(sessionID);
         explicitCoordinatorModels.delete(sessionID);
+        interruptedCoordinatorMessages.delete(sessionID);
         bindingDenials.delete(sessionID);
         sessionTaskIDs.delete(sessionID);
         recoverableWorkerChildren.delete(sessionID);
@@ -4780,17 +4856,17 @@ export const SortieDogsPlugin = async (input, options) => {
         parallelArtifactOperations.delete(sessionID);
         clearSessionLinks(sessionID);
     }
-    async function inspectSuccessfulRead(input) {
-        if (input.tool.toLowerCase() !== "read" || input.sessionID === undefined)
+    async function inspectSuccessfulRead(toolInput) {
+        if (toolInput.tool.toLowerCase() !== "read" || toolInput.sessionID === undefined)
             return;
-        if (activeSessionStatus(input.sessionID) !== "active" || !isRecord(input.args))
+        if (activeSessionStatus(toolInput.sessionID) !== "active" || !isRecord(toolInput.args))
             return;
-        const path = input.args.filePath;
+        const path = toolInput.args.filePath;
         if (typeof path !== "string" || path.length === 0)
             return;
-        const absolutePath = resolve(path);
-        const key = `${input.sessionID}\u0000${absolutePath}`;
-        const operation = inspect(path, input.sessionID).then(() => undefined);
+        const absolutePath = isAbsolute(path) ? resolve(path) : resolve(input.worktree ?? input.directory, path);
+        const key = `${toolInput.sessionID}\u0000${absolutePath}`;
+        const operation = inspectionOperations.get(key) ?? inspect(absolutePath, toolInput.sessionID).then(() => undefined);
         inspectionOperations.set(key, operation);
         try {
             await operation;
@@ -5219,10 +5295,33 @@ export const SortieDogsPlugin = async (input, options) => {
                 description: "Report handoff and operation manifest contract defects before dispatch. Read-only: it never inspects, binds, or authorizes.",
                 args: {
                     handoff_path: defineTool.schema.string(),
+                    task_prompt: optionalString(),
                 },
-                async execute(args) {
+                async execute(args, context) {
                     try {
                         await inspect(args.handoff_path, undefined, { report: true });
+                        if (typeof args.task_prompt === "string") {
+                            const state = await currentGoal(context.sessionID);
+                            const effectivePrompt = await resolveGoalPrompt(args.task_prompt);
+                            const resolved = resolveGoalDeclaration(state, effectivePrompt);
+                            if (resolved.defects.length > 0)
+                                return JSON.stringify({ status: "defective",
+                                    defects: resolved.defects.slice(0, CONTRACT_DEFECTS.limit),
+                                    remedy: "Correct the Task declaration before dispatch. Budget-only changes may omit the already accepted goal fields." });
+                            const units = Number(handoffValue(handoffEntries(effectivePrompt), ["goal_budget_units"]));
+                            const proposed = Number.isSafeInteger(units) && units > 0 && units >= state.consumed_units
+                                ? units : state.budget?.max_units ?? null;
+                            const approvalAvailable = goalDeclarationAuthority.get(context.sessionID) === state.latest_user_message_id &&
+                                state.latest_user_message_id !== null;
+                            return JSON.stringify({ status: "ok", defects: [], goal: {
+                                    goal_id: state.goal_id, revision: state.revision, inherited: resolved.inherited,
+                                    max_units: state.budget?.max_units ?? null, proposed_max_units: proposed,
+                                    consumed_units: state.consumed_units, reserved_units: state.outstanding_reservations.length,
+                                    remaining_units: state.budget === null ? null : Math.max(0, state.budget.max_units - state.consumed_units - state.outstanding_reservations.length),
+                                    proposed_remaining_units: proposed === null ? null : Math.max(0, proposed - state.consumed_units - state.outstanding_reservations.length),
+                                    budget_revision_requires_approval: proposed !== state.budget?.max_units && !approvalAvailable,
+                                } });
+                        }
                         return JSON.stringify({ status: "ok", defects: [] });
                     }
                     catch (error) {
@@ -5420,12 +5519,14 @@ export const SortieDogsPlugin = async (input, options) => {
         "experimental.text.complete": async (textInput, textOutput) => {
             if (pendingRealGoalTurns.has(textInput.sessionID))
                 await recoverPendingRealGoalTurn(textInput.sessionID);
+            const coordinatorReport = isCoordinatorSession(textInput.sessionID) || await recoverCoordinatorRoot(textInput.sessionID);
             if (fastLane.manualCompactionForbidden(textInput.sessionID)) {
                 textOutput.text = textOutput.text
                     .replaceAll(ROLLOVER_MARKER, "")
                     .replaceAll(CONTINUATION_MARKER, "")
                     .trimEnd();
             }
+            textOutput.text = await preserveActiveGoalContinuation(textInput.sessionID, textOutput.text, textInput.messageID);
             const runOutcome = terminalRunOutcome(textOutput.text);
             const terminal = runOutcome === undefined || !isCoordinatorSession(textInput.sessionID)
                 ? undefined
@@ -5442,7 +5543,7 @@ export const SortieDogsPlugin = async (input, options) => {
             if (runOutcome === "DONE" && terminal?.delivery === "failed") {
                 textOutput.text = replaceTerminalStatus(textOutput.text, "status: INTERRUPTED — durable delivery failed");
             }
-            if (runOutcome !== undefined && isCoordinatorSession(textInput.sessionID)) {
+            if (coordinatorReport && (runOutcome !== undefined || /<summary\b[^>]*>\s*Evidence\b/iu.test(textOutput.text))) {
                 textOutput.text = sanitizeTerminalReport(textOutput.text);
             }
             if ((isCoordinatorSession(textInput.sessionID) || await recoverCoordinatorRoot(textInput.sessionID)) &&
@@ -5561,6 +5662,8 @@ export const SortieDogsPlugin = async (input, options) => {
                 }
                 const coordinatorOrigin = parentID === undefined && requestedCoordinator;
                 if (coordinatorOrigin) {
+                    if (!synthetic)
+                        interruptedCoordinatorMessages.delete(chatInput.sessionID);
                     if (synthetic) {
                         if (messageID === undefined)
                             throw new Error("SORTIE_GOAL_CONTROL_DENIED: receiving-message-id-required");
@@ -5987,6 +6090,18 @@ export const SortieDogsPlugin = async (input, options) => {
             }
         },
         "tool.execute.before": async (toolInput, output) => {
+            // Publish the inspection at Read admission, before a concurrently scheduled bind can
+            // observe an empty cache. Bind joins this same host validation rather than asking the
+            // coordinator to launch another worker round merely for tool scheduling.
+            if (toolInput.tool.toLowerCase() === "read" && activeSessionStatus(toolInput.sessionID) === "active" &&
+                isRecord(output.args) && typeof output.args.filePath === "string") {
+                const path = output.args.filePath;
+                const absolutePath = isAbsolute(path) ? resolve(path) : resolve(input.worktree ?? input.directory, path);
+                const key = `${toolInput.sessionID}\u0000${absolutePath}`;
+                const operation = inspect(absolutePath, toolInput.sessionID).then(() => undefined);
+                inspectionOperations.set(key, operation);
+                void operation.catch(() => undefined); // The after hook/bind consumes the actual outcome.
+            }
             if (childLifecycles.get(toolInput.sessionID)?.stopping && toolInput.tool !== "sortie_release_write_gate") {
                 throw new Error("child-cancellation-in-progress");
             }
@@ -6161,12 +6276,12 @@ export const SortieDogsPlugin = async (input, options) => {
                         const resumeDeltaPresent = resumeDeltas.length === 1 && hasResumeContractShape(contractPrompt);
                         const contractRedefinitions = [
                             ...taskValues(contractPrompt, [
-                                "role", "project_root", "projectroot", "source_manifest", "sourcemanifest",
+                                "project_root", "projectroot", "source_manifest", "sourcemanifest",
                                 "acceptance", "validation", "validation_history", "validation_attempts", "scout",
                                 "known_facts", "known_paths", "relevant_constraints", "preserve",
                                 "parallel_group", "parallel_unit", "parallel_units",
                             ]),
-                            ...roleTokenValues(contractPrompt),
+                            ...taskValues(contractPrompt, ["role"]).filter(role => !TASK_ROLES.has(role)),
                         ];
                         if (modes.length !== 0 && !resume) {
                             throw new HandoffDeniedError("contract-invalid", "<worker-dispatch>", {
@@ -6516,6 +6631,7 @@ export const SortieDogsPlugin = async (input, options) => {
             if (eventSessionID === undefined)
                 return;
             if (event.type === "message.updated" && info !== undefined) {
+                rememberCoordinatorInterruption(eventSessionID, info);
                 await acceptPersistedRealGoalEvent(eventSessionID, info);
             }
             if (pendingRealGoalTurns.has(eventSessionID) &&
@@ -6549,8 +6665,9 @@ export const SortieDogsPlugin = async (input, options) => {
                     while (completedCoordinatorMessages.size > ACTIVE_SESSION_CACHE.maximum) {
                         completedCoordinatorMessages.delete(completedCoordinatorMessages.values().next().value);
                     }
-                    await terminalGoalFromHostText(eventSessionID, text);
-                    await completeContinuationText(eventSessionID, text, false);
+                    const effectiveText = await preserveActiveGoalContinuation(eventSessionID, text, eventPart.messageID);
+                    await terminalGoalFromHostText(eventSessionID, effectiveText);
+                    await completeContinuationText(eventSessionID, effectiveText, false);
                     return;
                 }
                 completedCoordinatorParts.delete(eventPart.id);
@@ -6583,8 +6700,9 @@ export const SortieDogsPlugin = async (input, options) => {
                     if (text === undefined)
                         completedCoordinatorMessages.delete(info.id);
                     else {
-                        await terminalGoalFromHostText(eventSessionID, text);
-                        await completeContinuationText(eventSessionID, text);
+                        const effectiveText = await preserveActiveGoalContinuation(eventSessionID, text, info.id);
+                        await terminalGoalFromHostText(eventSessionID, effectiveText);
+                        await completeContinuationText(eventSessionID, effectiveText);
                     }
                 }
                 catch {

package/dist/plugin/run-metrics.js

@@ -334,6 +334,8 @@ function duration(milliseconds) {
     if (seconds < 60)
         return `${seconds}s`;
     const minutes = Math.floor(seconds / 60);
+    if (minutes >= 60)
+        return `${Math.floor(minutes / 60)}h ${String(minutes % 60).padStart(2, "0")}m ${String(seconds % 60).padStart(2, "0")}s`;
     return `${minutes}m ${String(seconds % 60).padStart(2, "0")}s`;
 }
 function metricText(metric, render) {
@@ -348,15 +350,16 @@ export function formatSortieResult(result) {
         : result.mission.status === "INTERRUPTED" ? "中断（未完了）"
             : result.mission.status === "EXTERNAL_BLOCKER" ? "外部要因で未完了"
                 : "ユーザー判断待ち（未完了）";
-    return [
-        "**🐾 SORTIE DOGS — 帰還報告**",
-        `**⚡ 時間:** ${metricText(result.speed.goal_wall_ms, duration)}`,
-        `**🪙 使用量:** ${metricText(result.cost.total_tokens, (value) => `${value.toLocaleString("ja-JP")} tokens`)} · host推定額 ${metricText(result.cost.cost_usd, (value) => `$${value.toFixed(4)}`)}（実課金換算なし）`,
+    const color = result.mission.status === "COMPLETED" ? "🟢" : result.mission.status === "EXTERNAL_BLOCKER" ? "🔴" : "🟡";
+    const body = [
+        `**⚡ 任務経過（待機含む）:** **${metricText(result.speed.goal_wall_ms, duration)}**`,
+        `**🪙 使用量:** **${metricText(result.cost.total_tokens, (value) => `${value.toLocaleString("ja-JP")} tokens`)}** · host推定額 ${metricText(result.cost.cost_usd, (value) => `$${value.toFixed(4)}`)}（実課金換算なし）`,
         ...renderDebrief(result.debrief),
-        `**🛡 達成:** ${achievement} · 達成条件 ${criteria}`,
+        `**🛡 達成:** ${color} **${achievement}** · 達成条件 **${criteria}**`,
         "*最終応答生成前の計測*",
         ...renderCareer(result.career),
     ].join("\n\n");
+    return `<details>\n<summary><strong>🐾 SORTIE DOGS — 帰還報告｜${color} ${achievement}</strong></summary>\n\n${body}\n\n</details>`;
 }
 export function formatRunMetrics(metrics) {
     const elapsed = metrics.durationMilliseconds === undefined ? "duration unavailable" : `${duration(metrics.durationMilliseconds)} wall-clock`;
@@ -370,7 +373,16 @@ export function formatRunMetrics(metrics) {
 function topLevelLines(text) {
     const lines = [];
     let fence;
+    let disclosureDepth = 0;
     for (const [index, line] of text.split(/\r?\n/u).entries()) {
+        if (fence === undefined) {
+            const opens = (line.match(/<details\b[^>]*>/giu) ?? []).length;
+            const closes = (line.match(/<\/details>/giu) ?? []).length;
+            const inside = disclosureDepth > 0 || opens > 0;
+            disclosureDepth = Math.max(0, disclosureDepth + opens - closes);
+            if (inside)
+                continue;
+        }
         if (fence === undefined) {
             const opener = /^[ \t]*(`{3,}|~{3,})/u.exec(line)?.[1];
             if (opener === undefined) {
@@ -398,13 +410,11 @@ function terminalCheckpoint(text) {
         const explicit = normalized === "DONE" || normalized === "INTERRUPTED" || normalized === "BLOCKED" || normalized === "NEED_DECISION"
             ? normalized
             : undefined;
-        const outcome = explicit ?? conclusionStatusAlias(line) ??
-            (/^✅[ \t]+\*\*DONE\*\*/u.test(line) ? "DONE" :
-                /^⚠️[ \t]+\*\*INTERRUPTED\*\*/u.test(line) ? "INTERRUPTED" :
-                    /^⛔[ \t]+\*\*BLOCKED\*\*/u.test(line) ? "BLOCKED" :
-                        /^❓[ \t]+\*\*NEED_DECISION\*\*/u.test(line) ? "NEED_DECISION" : undefined);
+        // The status carries meaning; a decorative icon must not suppress terminal rendering.
+        const decorated = /^(?:\p{Extended_Pictographic}\uFE0F?[ \t]+)?\*\*(DONE|INTERRUPTED|BLOCKED|NEED_DECISION)\*\*(?=[ \t]|$)/u.exec(line)?.[1];
+        const outcome = explicit ?? conclusionStatusAlias(line) ?? decorated;
         return outcome === "DONE" || outcome === "INTERRUPTED" || outcome === "BLOCKED" || outcome === "NEED_DECISION"
-            ? { index, outcome }
+            ? { index, outcome: outcome }
             : undefined;
     })();
     return checkpoint;
@@ -437,12 +447,20 @@ export function replaceDoneTerminalStatus(text, replacement) {
 }
 export function sanitizeTerminalReport(text) {
     const newline = text.includes("\r\n") ? "\r\n" : "\n";
-    const internal = /\b(?:evidence_refs?|manifest|raw|raw_status|reason_code|goal_control|TRUE_BLOCKER)\s*:|\bEvidence\b/iu;
-    return text.replace(/^[ \t]*(`{3,}|~{3,})[^\r\n]*\r?\n([\s\S]*?)^[ \t]*\1[ \t]*$/gimu, (block, _fence, body) => internal.test(body) ? "" : block)
-        .replace(/<details\b[^>]*>[\s\S]*?<\/details>/giu, "")
+    const preserved = [];
+    const keep = (block) => `\0SORTIE_PROSE_${preserved.push(block) - 1}\0`;
+    return text.replace(/<details\b[^>]*>[\s\S]*?<\/details>/giu, block => {
+        const summary = /<summary\b[^>]*>([\s\S]*?)<\/summary>/iu.exec(block)?.[1]?.replace(/<[^>]*>/gu, "").trim() ?? "";
+        return /^(?:Evidence\b|🐾\s*SORTIE DOGS\s*[—-]\s*帰還報告)/iu.test(summary) ? "" : keep(block);
+    })
+        .replace(/^[ \t]*(`{3,}|~{3,})[^\r\n]*\r?\n[\s\S]*?^[ \t]*\1[ \t]*$/gimu, (block, _fence, offset, source) => {
+        const preceding = source.slice(0, offset).trimEnd().split(/\r?\n\s*\r?\n/u).at(-1) ?? "";
+        return /^[ \t]*(?:#{1,6}[ \t]*)?\**Evidence\**[ \t]*:/imu.test(preceding) ? "" : keep(block);
+    })
         .split(/\r?\n/u)
         .filter((line) => !/^\s*(?:(?:#{1,6}\s*)?\**Evidence\**\s*:|(?:TRUE_BLOCKER|goal_control|evidence_refs?|reason_code|raw|raw_status|manifest)\s*:)/iu.test(line))
         .join(newline)
+        .replace(/\0SORTIE_PROSE_(\d+)\0/gu, (_token, index) => preserved[Number(index)])
         .trimEnd();
 }
 export function insertRunMetrics(text, metrics) {
@@ -465,8 +483,24 @@ export function insertSortieResult(text, result) {
     // A title in model text is not trusted evidence. Replace existing cards, including legacy cards.
     const newline = visible.includes("\r\n") ? "\r\n" : "\n";
     const lines = visible.split(/\r?\n/u);
-    const cardLines = new Set(topLevelLines(visible).filter(({ index, line }) => index > checkpoint.index &&
-        /^(?:\*\*(?:Sortie Result|🐾 SORTIE DOGS — 帰還報告|📜 PACK RECORD|↳\*\*|(?:Speed|Cost|達成|⚡ 時間|🪙 使用量|🐕 出撃隊|モデル別token内訳|実行重複率|🛡 達成|確認|🏅 今回の戦績|戦績|初回完遂|累積使用量|累積モデル|累積時間|累積実行重複率|保存範囲|🎖 隊の称号):)|\*最終応答生成前の計測\*)/u.test(line)).map(({ index }) => index));
+    const cardLines = new Set();
+    let legacyCard = false;
+    let previousIndex = checkpoint.index;
+    for (const { index, line } of topLevelLines(visible)) {
+        if (index <= checkpoint.index)
+            continue;
+        if (index > previousIndex + 1)
+            legacyCard = false;
+        previousIndex = index;
+        if (/^\*\*(?:Sortie Result|🐾 SORTIE DOGS — 帰還報告|📜 PACK RECORD)/u.test(line))
+            legacyCard = true;
+        if (legacyCard && (line.trim() === "" ||
+            /^(?:\*\*(?:Sortie Result|🐾 SORTIE DOGS — 帰還報告|📜 PACK RECORD|↳\*\*|(?:Speed|Cost|達成|⚡ 時間|🪙 使用量|🐕 出撃隊|モデル別token内訳|実行重複率|🛡 達成|確認|🏅 今回の戦績|戦績|初回完遂|累積使用量|累積モデル|累積時間|累積実行重複率|保存範囲|🎖 隊の称号):)|\*最終応答生成前の計測\*)/u.test(line))) {
+            cardLines.add(index);
+        }
+        else
+            legacyCard = false;
+    }
     const cleaned = lines.filter((_, index) => !cardLines.has(index));
     while (cleaned[checkpoint.index + 1] === "")
         cleaned.splice(checkpoint.index + 1, 1);
@@ -475,7 +509,7 @@ export function insertSortieResult(text, result) {
         card = formatSortieResult(result);
     }
     catch {
-        card = "**🐾 SORTIE DOGS — 帰還報告**\n**確認:** 表示集計を取得できません。任務結果は先頭の状態を参照。";
+        card = "<details>\n<summary><strong>🐾 SORTIE DOGS — 帰還報告</strong></summary>\n\n**確認:** 表示集計を取得できません。任務結果は先頭の状態を参照。\n\n</details>";
     }
     cleaned.splice(checkpoint.index + 1, 0, "", card, "");
     return cleaned.join(newline).trimEnd();

package/dist/plugin/sortie-debrief.d.ts

@@ -17,11 +17,13 @@ export interface DebriefSession {
     readonly reviews: Array<{
         at: number;
         status: "PASS" | "FAIL" | "WAIVED";
+        source?: "reviewer" | "controller";
     }>;
     readonly tasks: string[];
     complete: boolean;
     timingComplete: boolean;
     failed: boolean;
+    lastPossibleMutation?: number;
 }
 export interface DebriefObservation {
     readonly complete: boolean;
@@ -40,6 +42,8 @@ export interface Debrief {
     }[] | null;
     readonly validation: "PASS" | "FAIL" | "未確認";
     readonly review: "PASS" | "FAIL" | "WAIVED" | "未確認";
+    readonly reviewSource?: "reviewer" | "controller";
+    readonly notes?: readonly string[];
     readonly traits: readonly ("連携作戦" | "修正から復帰" | "一発完遂")[];
     readonly firstPassEligible?: boolean;
     readonly overlap?: {

package/dist/plugin/sortie-debrief.js

@@ -1,8 +1,9 @@
 import { createHash } from "node:crypto";
 import { DEDICATED_WORKER_ROLES, LUNA_FABRIC_WORKER_ROLE } from "./model-routing.js";
+import { normalizeCommand } from "./gate.js";
 const object = (value) => value !== null && typeof value === "object" ? value : undefined;
 const timeValue = (value) => typeof value === "number" && Number.isFinite(value) && value >= 0 ? value : undefined;
-const fingerprint = (value) => createHash("sha256").update(value).digest("hex");
+const fingerprint = (value) => createHash("sha256").update(normalizeCommand(value)).digest("hex");
 const workerRoles = new Set([...DEDICATED_WORKER_ROLES, LUNA_FABRIC_WORKER_ROLE]);
 function unionDuration(spans) {
     let end = -1, total = 0;
@@ -34,14 +35,11 @@ export function observeDebriefSession(id, root, messages, window) {
             session.timingComplete = false;
             continue;
         }
-        if (!within(span, window)) {
-            if (window !== undefined && span.start < window.end && span.end > window.start) {
-                session.complete = false;
-                session.timingComplete = false;
-            }
+        // Match the token collector's completed-message window. A message can start just before
+        // goal.accepted; clip its timing rather than invalidating every independent metric.
+        if (window !== undefined && (span.end < window.start || span.end > window.end))
             continue;
-        }
-        session.spans.push(span);
+        session.spans.push(window === undefined ? span : { start: Math.max(span.start, window.start), end: span.end });
         if (info.error !== undefined)
             session.failed = true;
         for (const raw of Array.isArray(message.parts) ? message.parts : []) {
@@ -89,8 +87,20 @@ export function observeDebriefSession(id, root, messages, window) {
                     else
                         session.complete = false;
                 }
+                if (args?.subagent_type === "dog-reviewer" && typeof state.output === "string") {
+                    const body = /<task_result>([\s\S]*?)<\/task_result>/u.exec(state.output)?.[1] ?? state.output;
+                    const lines = body.replace(/^[ \t]*(`{3,}|~{3,})[^\r\n]*\r?\n[\s\S]*?^[ \t]*\1[ \t]*$/gmu, "")
+                        .trim().split(/\r?\n/u);
+                    const verdicts = [lines[0], lines.at(-1)].map(line => line?.trim().match(/^(?:\*\*)?(?:(?:verdict|結論|判定)\s*:\s*)?(PASS|FAIL|MUST_FIX)(?:\*\*)?$/iu)?.[1]?.toUpperCase());
+                    const verdict = verdicts.some(value => value === "FAIL" || value === "MUST_FIX") || /^FINDING\b/iu.test(lines[0] ?? "")
+                        ? "FAIL" : verdicts.includes("PASS") ? "PASS" : undefined;
+                    if (verdict !== undefined)
+                        session.reviews.push({ at: interval.end,
+                            status: verdict === "PASS" ? "PASS" : "FAIL", source: "reviewer" });
+                }
             }
             if (["edit", "write", "apply_patch"].includes(tool)) {
+                session.lastPossibleMutation = Math.max(session.lastPossibleMutation ?? 0, interval.end);
                 const metadata = object(state.metadata);
                 const changed = typeof metadata?.diff === "string" && /^[+-](?![+-])/mu.test(metadata.diff) ||
                     Array.isArray(metadata?.files) && metadata.files.some((file) => {
@@ -138,15 +148,14 @@ export function observeDebriefSession(id, root, messages, window) {
     return session;
 }
 export function buildDebrief(receipt, contract, observation, records) {
-    const empty = { pack: null, mix: null, validation: "未確認", review: "未確認", traits: [] };
-    if (observation === undefined)
-        return empty;
+    observation ??= { complete: false, sessions: [], window: undefined };
     const sessions = observation.sessions;
     const complete = observation.complete && sessions.every((session) => session.complete);
     const timingComplete = observation.complete && sessions.every((session) => session.timingComplete);
-    const children = sessions.filter((session) => !session.root && session.spans.length > 0);
+    const children = sessions.filter((session) => !session.root &&
+        (session.spans.length > 0 || Object.keys(session.models).length > 0));
     const counts = new Map(), totals = new Map();
-    let usageComplete = timingComplete;
+    let usageComplete = observation.complete;
     for (const session of sessions) {
         for (const [model, tokens] of Object.entries(session.models)) {
             if (tokens === null || model === "未分類")
@@ -163,11 +172,25 @@ export function buildDebrief(receipt, contract, observation, records) {
     const total = [...totals.values()].reduce((sum, tokens) => sum + tokens, 0);
     const commands = new Set(contract?.criteria.flatMap((criterion) => criterion.validation_command === undefined ? [] : [fingerprint(criterion.validation_command)]));
     const checks = sessions.flatMap((session) => session.checks.filter((check) => commands.has(check.command))).sort((a, b) => a.end - b.end);
+    // Retain canonical proof even when an unrelated host tool lacks timing/metadata.
+    for (const { event } of records ?? []) {
+        if (event.kind !== "unit.settled" || event.goal_id !== receipt.goal_id)
+            continue;
+        for (const evidence of event.evidence ?? []) {
+            const command = fingerprint(evidence.execution.command.join(" "));
+            const start = Date.parse(evidence.execution.started_at), end = Date.parse(evidence.execution.ended_at);
+            if (receipt.evidence_refs.includes(evidence.evidence_id) && commands.has(command) &&
+                Number.isFinite(start) && Number.isFinite(end))
+                checks.push({ start, end, command,
+                    passed: evidence.execution.exit_code === 0 && evidence.execution.outcome === "pass" });
+        }
+    }
+    checks.sort((a, b) => a.end - b.end);
     const latest = new Map();
     for (const check of checks)
         latest.set(check.command, check);
     const validation = [...latest.values()].some((check) => !check.passed) ? "FAIL"
-        : complete && commands.size > 0 && latest.size === commands.size ? "PASS" : "未確認";
+        : commands.size > 0 && latest.size === commands.size ? "PASS" : "未確認";
     const reviews = sessions.flatMap((session) => session.reviews).sort((a, b) => a.at - b.at);
     const traits = [];
     const spans = children.flatMap((session) => session.spans.filter((span) => span.end > span.start)
@@ -218,9 +241,15 @@ export function buildDebrief(receipt, contract, observation, records) {
             (event.kind === "validation.admission" && event.decision === "DENY") || (event.kind === "validation.settled" && event.outcome !== "passed")) &&
         sessions.every((session) => !session.failed && new Set(session.checks.map((check) => check.command)).size === session.checks.length))
         traits.push("一発完遂");
-    return { pack: timingComplete ? [...counts].sort(([a], [b]) => a.localeCompare(b)).map(([model, count]) => ({ model, count })) : null,
+    const review = reviews.filter((entry) => !sessions.some((session) => (session.lastPossibleMutation ?? 0) > entry.at ||
+        session.mutations.some((edit) => edit.end > entry.at))).at(-1);
+    const packComplete = observation.complete && sessions.every(session => session.root || session.timingComplete || Object.keys(session.models).length > 0);
+    return { pack: packComplete ? [...counts].sort(([a], [b]) => a.localeCompare(b)).map(([model, count]) => ({ model, count })) : null,
         mix: usageComplete && total > 0 ? [...totals].sort(([a], [b]) => a.localeCompare(b)).map(([model, tokens]) => ({ model, tokens, percent: tokens / total * 100 })) : null,
-        validation, review: complete ? reviews.filter((review) => !sessions.some((session) => session.mutations.some((edit) => edit.end > review.at))).at(-1)?.status ?? "未確認" : "未確認", traits, firstPassEligible,
+        validation, review: review?.status ?? "未確認", reviewSource: review?.source ?? "controller", traits, firstPassEligible,
+        notes: [...(!observation.complete ? ["host履歴の取得が一部不足"] : []),
+            ...(!usageComplete ? ["モデルIDまたはusageの記録が不足"] : []),
+            ...(!timingComplete ? ["稼働区間の時刻が不足（token集計とは独立）"] : [])],
         ...(timingComplete ? { overlap: { workerMilliseconds: children.reduce((sum, child) => sum + unionDuration(child.spans), 0),
                 wallMilliseconds: unionDuration(spans) } } : {}) };
 }
@@ -239,13 +268,17 @@ export function renderDebrief(debrief) {
         const filled = Math.max(0, Math.min(10, Math.round(percent / 10)));
         return "█".repeat(filled) + "░".repeat(10 - filled);
     };
+    const status = (value) => value === "PASS" ? "🟢 **PASS**" : value === "FAIL" ? "🔴 **FAIL**"
+        : value === "WAIVED" ? "免除" : "未記録";
     return [
-        `**🐕 出撃隊:** ${pack === null ? "計測不可" : pack.length === 0 ? "出撃なし" : packVisible.map((entry) => `${label(entry.model)} ×${entry.count}`).join(" · ")}`,
-        `**モデル別token内訳:** ${mix === null ? "計測不可" : ""}`,
+        `**🐕 出撃隊:** ${pack === null ? "履歴未取得" : pack.length === 0 ? "出撃なし" : packVisible.map((entry) => `${label(entry.model)} **×${entry.count}**`).join(" · ")}`,
+        `**モデル別token内訳:** ${mix === null ? "usage未取得" : ""}`,
         ...visible.map((entry) => `**↳** ${label(entry.model)} \`${bars(entry.percent)}\` ${entry.percent.toFixed(1)}%`),
         `**実行重複率:** ${debrief?.overlap !== undefined && debrief.overlap.wallMilliseconds > 0
-            ? `${(debrief.overlap.workerMilliseconds / debrief.overlap.wallMilliseconds).toFixed(2)}×（worker区間・速度倍率ではありません）` : "計測不可"}`,
-        `**確認:** 対象検証 ${debrief?.validation ?? "未確認"} · 直近Review ${debrief?.review === "WAIVED" ? "免除" : debrief?.review ?? "未確認"}`,
+            ? `**${(debrief.overlap.workerMilliseconds / debrief.overlap.wallMilliseconds).toFixed(2)}×**（worker区間・速度倍率ではありません）`
+            : pack?.length === 0 ? "対象なし（出撃なし）" : "稼働区間の記録不足"}`,
+        `**確認:** 対象検証 ${status(debrief?.validation ?? "未確認")} · 直近Review ${status(debrief?.review ?? "未確認")}${debrief?.reviewSource === "reviewer" ? "（reviewer報告）" : ""}`,
+        ...(debrief?.notes?.length ? [`**計測範囲:** ${debrief.notes.join(" · ")}`] : []),
         ...(debrief?.traits.length ? [`**🏅 今回の戦績:** ${debrief.traits.join(" · ")}`] : []),
     ];
 }

package/dist/runtime-assets.d.ts

@@ -7,7 +7,7 @@ export interface RuntimeAsset {
 }
 export declare const runtimeAssets: readonly [{
     readonly name: "dog-coordinator";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-coordinator.md";
     readonly content: `---
 description: Canonical MkII coordinator packaged by Sortie-dogs
@@ -575,7 +575,8 @@ END_ONE_WORKER_EXECUTION_CLOSURE_FIXTURE
 
 Use a same-task resume only when the runtime denial explicitly returns resume_session=true for that
 exact child. A completed Task without that signal requires a fresh worker and full handoff. For an
-authorized same-task resume, retain the prior effective digest. Send the same task_id and only a
+authorized same-task resume, retain the prior effective digest. Optional role: blocker-resolution
+names the recovery action without replacing scope or acceptance. Send the same task_id and a
 resume_delta containing stale_paths, new_findings, the previous command exit/fingerprint, and
 next_action. Do not resend unchanged acceptance, role, validation, facts, constraints, manifests,
 or file content; the preserved values plus this delta form the effective digest.
@@ -854,9 +855,19 @@ acceptance-continuity parent_fingerprint. SORTIE_ACCEPTANCE_CONTINUITY_STATE is
 projection of the latest gate-accepted sequential unit; its next_sequential_parent_fingerprint is the
 only projected value for the next unit's parent when present.
 
-Declare goal_acceptance_fingerprint, delivery_intent, delivery_mode when explicitly selected,
-usable_path_established, controlled_change, and goal_budget_units in the first worker handoff produced
-from the current real-user turn. For each terminal criterion also declare goal_criterion_id, goal_target,
+A goal may be declared once in JSON and referenced by goal_declaration_path in Task, or embedded as
+ext["sortie-dogs/goal-declaration"] in its registered handoff. Shared defaults apply to each item in
+criteria, with explicit criterion overrides. The host supplies stable IDs and a fingerprint if omitted.
+Do not expand this definition into worker instructions; the host resolves it privately. Acceptance,
+oracle coverage, delivery choice, and validation remain explicit. Legacy flat input remains supported:
+declare goal_acceptance_fingerprint, delivery_intent, delivery_mode when explicitly selected,
+usable_path_established, controlled_change, and goal_budget_units when first accepting a goal or changing
+its acceptance contract. A same-goal continuation with only an approved budget change may send just
+goal_budget_units (or the changed time/cost budget) beside the ordinary Task handoff: the host inherits
+the accepted goal declaration. Do not reconstruct or repeat that declaration merely because the user
+resumed the session. Optional task_prompt on sortie_check_contract previews declaration inheritance
+and current/proposed capacity before dispatch without consuming approval or reserving a unit.
+For each new or revised terminal criterion also declare goal_criterion_id, goal_target,
 goal_entrypoint, goal_workload, goal_oracle_coverage, goal_build_boundary, goal_fixture, goal_proof_scope,
 goal_expected_outcome, and the exact goal_validation_command from the operation manifest. Use
 goal_source_binding: current-protected and goal_candidate_binding: current-protected when implementation
@@ -1399,23 +1410,33 @@ COMMIT_SCOPE_FIXTURE
     mismatch: commit rejected
 END_COMMIT_SCOPE_FIXTURE
 
+When the user asks what you are doing, why, or requests an explanation, answer the actual question
+with sufficient context, completed work, remaining work, and rationale. The terminal format below
+applies only to a task's terminal report, not to ordinary questions. Do not replace a substantive
+explanation with a status/Validation/Next template. Preserve explanatory paragraphs, code examples,
+and explanatory disclosure sections; a status annotation must not erase the user's answer.
+
 At each checkpoint and terminal return, preserve concise proof internally. The user-facing terminal
 return MUST begin with its conclusion: no plan, progress, assessment, Evidence heading, or preamble.
 Use exactly one of DONE, INTERRUPTED, BLOCKED, or NEED_DECISION with one status emoji and a short
 Japanese conclusion. Then render Japanese \u5909\u66F4\u70B9, \u78BA\u8A8D\u7D50\u679C, and \u6B21 paragraphs without bullets or decorative
-emoji. The plugin injects measured Speed, Cost, and \u9054\u6210 paragraphs in a Japanese mission debrief card,
+emoji. The plugin injects measured Speed, Cost, and \u9054\u6210 paragraphs in a collapsed Japanese mission debrief card,
 with one fixed icon per section, observed pack/model usage, validation/review, and evidence-backed traits.
 Its Markdown token bars and PACK RECORD summarize retained project goals, with coverage and team titles;
 they never imply lifetime history, XP, levels, unmeasured savings, or a leaderboard rank.
 Never write the card, its metrics, or its badges yourself. Do not estimate or fabricate them.
 Use \u4EFB\u52D9\u5B8C\u4E86 for DONE, \u4E2D\u65AD\u5E30\u9084\uFF08\u672A\u5B8C\u4E86\uFF09 for INTERRUPTED, \u5916\u90E8\u8981\u56E0\u3067\u5F85\u6A5F\uFF08\u672A\u5B8C\u4E86\uFF09 for BLOCKED,
 and \u6307\u793A\u5F85\u3061\uFF08\u672A\u5B8C\u4E86\uFF09 for NEED_DECISION; preserve the machine status token and first-line checkpoint.
-Never render a user-facing Evidence heading, <details> block, evidence reference, internal reason code,
+Never render a user-facing Evidence heading or Evidence details block, evidence reference, internal reason code,
 ledger key, or raw status. Keep ordered command/exit/fingerprint history, manifests, evidence refs,
 review proof, and terminal receipt append-only in their internal typed ledger and host logs. A concise
 \u78BA\u8A8D\u7D50\u679C may summarize PASS/FAIL without exposing those internal identifiers.
 An undeclared write or mutation must be reported as rejected, not performed. A locally repairable process or evidence defect is never a
 user question: repair it and continue in the same turn.
+An INTERRUPTED root conclusion requires one machine line: \`TRUE_INTERRUPTION: user: <condition>\`
+for an explicit user stop, or \`TRUE_INTERRUPTION: internal: <condition>\` for a true internal
+interruption. Never emit that line for a local process defect, step boundary, continuation request, or
+recoverable limit; keep those on the same-session autonomous continuation path.
 
 TERMINAL_STATUS_SEMANTICS_FIXTURE
     DONE: all accepted criteria proved complete; unmet or interrupted work forbidden
@@ -1426,13 +1447,15 @@ TERMINAL_STATUS_SEMANTICS_FIXTURE
     status_icons: DONE=\u2705 | INTERRUPTED=\u26A0\uFE0F | BLOCKED=\u26D4 | NEED_DECISION=\u2753
     quality_gate_fail: validation evidence + autonomous non-adoption decision -> DONE; release remains unperformed
     process_defect: gate | routing | handoff | local tool defect -> autonomous repair; never terminal BLOCKED
+    interruption_marker: TRUE_INTERRUPTION: user: <condition> | TRUE_INTERRUPTION: internal: <condition>
+    continuation_not_interruption: local process defect | step boundary | continuation request | recoverable limit
 END_TERMINAL_STATUS_SEMANTICS_FIXTURE
 
 RUNTIME_ASSET_VERSION_SYNC_FIXTURE
-    runtime_version: 0.3.81-mission-debrief-v1
+    runtime_version: 0.3.84-dispatch-recovery-v1
     shared_marker: src/asset-version.ts
-    packaged_expectation: test/plugin-loader.test.ts uses 0.3.81-mission-debrief-v1
-    initialize_expectation: test/initialize.test.ts uses 0.3.81-mission-debrief-v1
+    packaged_expectation: test/plugin-loader.test.ts uses 0.3.84-dispatch-recovery-v1
+    initialize_expectation: test/initialize.test.ts uses 0.3.84-dispatch-recovery-v1
     rule: runtime asset versions, shared marker, packaged expectation, and initialize expectation change together
 END_RUNTIME_ASSET_VERSION_SYNC_FIXTURE
 
@@ -1455,32 +1478,32 @@ END_INTERNAL_TERMINAL_PROOF_FIXTURE
 `;
 }, {
     readonly name: "dog-worker";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-worker.md";
     readonly content: string;
 }, {
     readonly name: "dog-luna-worker";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-luna-worker.md";
     readonly content: string;
 }, {
     readonly name: "dog-scout";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-scout.md";
     readonly content: "---\ndescription: Bounded evidence scout for dog-coordinator\nmode: subagent\nsteps: 8\npermission:\n  bash: deny\n  webfetch: deny\n  task: deny\n  question: deny\n  glob: deny\n  grep: deny\n  edit: deny\n  list: deny\n  write: deny\n  patch: deny\ntools:\n  bash: false\n  webfetch: false\n  task: false\n  question: false\n  glob: false\n  grep: false\n  edit: false\n  list: false\n  write: false\n  patch: false\n---\n# dog-scout\n\nAccept one concrete missing_evidence_code: manifest, validation, or owner-risk. Accept only an\nexplicit absolute project_root and a known_paths list of at most four paths from dog-coordinator.\nResolve only that evidence key from those paths under project_root; never resolve a path against the\nsession directory. Use Read only, with at most 120 lines and no more than one read per supplied path.\nDo not resolve a second key, explore, invoke another tool, retry, edit, stage, commit, or become user-facing.\n\nWhen project_root is missing, or a supplied path does not resolve under it, or a resolved path is\nunreadable, report that dispatch defect as the facts for the requested key and name the exact paths.\nDo not retry, guess another root, or answer from an unread path.\n\nReturn exactly one concise JSON object of at most 800 characters with exactly these keys:\nmissing_evidence_code, facts, evidence_paths, risks. Use no Markdown, code fence, commentary, or raw log. Return it only\nto dog-coordinator. Write the facts and risks prose in the language the dispatch uses for its own\nprose; keep the keys, paths, commands, and identifiers verbatim.\n";
 }, {
     readonly name: "dog-reviewer";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-reviewer.md";
     readonly content: "---\ndescription: Independent source reviewer for dog-coordinator\nmode: subagent\npermission:\n  bash: deny\n  webfetch: deny\n  task: deny\n  question: deny\n  glob: deny\n  grep: deny\n  edit: deny\n  list: deny\n  write: deny\n  patch: deny\n  read: deny\ntools:\n  bash: false\n  webfetch: false\n  task: false\n  question: false\n  glob: false\n  grep: false\n  edit: false\n  list: false\n  write: false\n  patch: false\n  read: false\n---\n# dog-reviewer\n\nAccept only one bounded SourceReview request from dog-coordinator after canonical\nvalidation for one high-risk candidate. Review only the supplied acceptance criteria, exact\nmanifest, changedLogicSummary, supplied changed-code excerpts, and validation evidence. Confirm every acceptance item explicitly\nmaps to at least one changedLogicSummary entry and assess that changed logic against the mapped\nacceptance item. Missing or incomplete coverage is a concrete finding, never PASS.\nRequire one indexed acceptance[i] -> changedLogicSummary[j] mapping line per acceptance item and\nreject a missing index or unequal mapping count before assessing the changed logic.\nDo not request raw logs or full source files, review low-risk candidates, expand scope, or dispatch\nanother agent.\nTreat those supplied fields as the complete bounded SourceReview artifact; use only that artifact and invoke no tools.\nDo not infer that a branch or exemption is absent from source because a prose summary omits it.\nIf the supplied excerpts do not establish a claim, report an evidence gap and request the exact\nbranch/helper excerpt in the next artifact; do not prescribe a source fix for an unproven defect.\n\nReturn one concise PASS or concrete-finding response only to dog-coordinator before the\ncoordinator commit. Write every finding, evidence, and required-fix sentence in the language the\nsupplied artifact uses for its own prose, one statement per line, and keep verdict values,\nidentifiers, paths, and commands verbatim. Do not implement, remediate, resolve blockers, edit,\nstage, commit, or become user-facing. Remain host-routed: do not require or identify a provider, vendor, model, variant,\nor transport.\n";
 }, {
     readonly name: "dog-advisor";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "agent/dog-advisor.md";
     readonly content: "---\ndescription: Focused technical advisor for dog-coordinator\nmode: subagent\npermission:\n  bash: deny\n  webfetch: deny\n  task: deny\n  question: deny\n  glob: deny\n  grep: deny\n  edit: deny\n  list: deny\n  write: deny\n  patch: deny\n  read: deny\ntools:\n  bash: false\n  webfetch: false\n  task: false\n  question: false\n  glob: false\n  grep: false\n  edit: false\n  list: false\n  write: false\n  patch: false\n  read: false\n---\n# dog-advisor\n\nAccept only one bounded Strategy request from dog-coordinator for one candidate and one focused\nquestion. Use only the supplied acceptance criteria, exact manifest, constraints, and concise\nevidence. Do not request raw logs or full source files, expand scope, or dispatch another agent.\nTreat those supplied fields as the complete bounded Strategy artifact; use only that artifact and invoke no tools.\nReject every SourceReview request and return the rejection only to dog-coordinator; SourceReview is\ndog-reviewer-only work.\n\nReturn concise options and one recommendation only to dog-coordinator. Write every option,\nrecommendation, and consideration in the language the supplied request uses for its own prose, one\nstatement per line, and keep identifiers, paths, and commands verbatim. Do not perform\nSourceReview, implement, remediate, resolve blockers, edit, stage, commit, or become user-facing.\nImplementation remains dog-worker work. Remain host-routed: do not require or identify a\nprovider, vendor, model, variant, or transport.\n";
 }, {
     readonly name: "sortie";
-    readonly version: "0.3.81-mission-debrief-v1";
+    readonly version: "0.3.84-dispatch-recovery-v1";
     readonly installPath: "command/sortie.md";
     readonly content: "---\ndescription: Start the canonical Sortie-dogs MkII workflow\nagent: dog-coordinator\n---\nRequest: $ARGUMENTS\n\n1. If $ARGUMENTS is empty, request task context and stop; give project init guidance first.\n2. Do not preflight installed runtime assets. The plugin reports version skew without adding model\n   turns; proceed from task evidence and project instructions.\n3. On restart or re-entry, reconstruct context from project-local durable artifacts and the\n   latest bounded handoff or checkpoint. Preserve both manifests and ordered validation history;\n   resume the same task through dog-coordinator with only the required delta.\n4. Otherwise transfer request and project context to dog-coordinator. Frontmatter is the single coordinator\n   transfer; never route a worker to the user.\n";
 }];

package/dist/runtime-assets.js

@@ -1,5 +1,5 @@
 import { GOAL_DECLARATION_FORMAT } from "./core/goal-declaration-format.js";
-const ASSET_VERSION = "0.3.81-mission-debrief-v1";
+const ASSET_VERSION = "0.3.84-dispatch-recovery-v1";
 // Kept local so source-mode CLI execution does not load the plugin graph.
 const BACKLOG_DRAIN_CAPABILITY = "sortie_enable_backlog_drain";
 const PARALLEL_PREPARE_CAPABILITY = "sortie_prepare_parallel_dispatch";
@@ -782,7 +782,8 @@ END_ONE_WORKER_EXECUTION_CLOSURE_FIXTURE
 
 Use a same-task resume only when the runtime denial explicitly returns resume_session=true for that
 exact child. A completed Task without that signal requires a fresh worker and full handoff. For an
-authorized same-task resume, retain the prior effective digest. Send the same task_id and only a
+authorized same-task resume, retain the prior effective digest. Optional role: blocker-resolution
+names the recovery action without replacing scope or acceptance. Send the same task_id and a
 resume_delta containing stale_paths, new_findings, the previous command exit/fingerprint, and
 next_action. Do not resend unchanged acceptance, role, validation, facts, constraints, manifests,
 or file content; the preserved values plus this delta form the effective digest.
@@ -1061,9 +1062,19 @@ acceptance-continuity parent_fingerprint. SORTIE_ACCEPTANCE_CONTINUITY_STATE is
 projection of the latest gate-accepted sequential unit; its next_sequential_parent_fingerprint is the
 only projected value for the next unit's parent when present.
 
-Declare goal_acceptance_fingerprint, delivery_intent, delivery_mode when explicitly selected,
-usable_path_established, controlled_change, and goal_budget_units in the first worker handoff produced
-from the current real-user turn. For each terminal criterion also declare goal_criterion_id, goal_target,
+A goal may be declared once in JSON and referenced by goal_declaration_path in Task, or embedded as
+ext["sortie-dogs/goal-declaration"] in its registered handoff. Shared defaults apply to each item in
+criteria, with explicit criterion overrides. The host supplies stable IDs and a fingerprint if omitted.
+Do not expand this definition into worker instructions; the host resolves it privately. Acceptance,
+oracle coverage, delivery choice, and validation remain explicit. Legacy flat input remains supported:
+declare goal_acceptance_fingerprint, delivery_intent, delivery_mode when explicitly selected,
+usable_path_established, controlled_change, and goal_budget_units when first accepting a goal or changing
+its acceptance contract. A same-goal continuation with only an approved budget change may send just
+goal_budget_units (or the changed time/cost budget) beside the ordinary Task handoff: the host inherits
+the accepted goal declaration. Do not reconstruct or repeat that declaration merely because the user
+resumed the session. Optional task_prompt on sortie_check_contract previews declaration inheritance
+and current/proposed capacity before dispatch without consuming approval or reserving a unit.
+For each new or revised terminal criterion also declare goal_criterion_id, goal_target,
 goal_entrypoint, goal_workload, goal_oracle_coverage, goal_build_boundary, goal_fixture, goal_proof_scope,
 goal_expected_outcome, and the exact goal_validation_command from the operation manifest. Use
 goal_source_binding: current-protected and goal_candidate_binding: current-protected when implementation
@@ -1606,23 +1617,33 @@ COMMIT_SCOPE_FIXTURE
     mismatch: commit rejected
 END_COMMIT_SCOPE_FIXTURE
 
+When the user asks what you are doing, why, or requests an explanation, answer the actual question
+with sufficient context, completed work, remaining work, and rationale. The terminal format below
+applies only to a task's terminal report, not to ordinary questions. Do not replace a substantive
+explanation with a status/Validation/Next template. Preserve explanatory paragraphs, code examples,
+and explanatory disclosure sections; a status annotation must not erase the user's answer.
+
 At each checkpoint and terminal return, preserve concise proof internally. The user-facing terminal
 return MUST begin with its conclusion: no plan, progress, assessment, Evidence heading, or preamble.
 Use exactly one of DONE, INTERRUPTED, BLOCKED, or NEED_DECISION with one status emoji and a short
 Japanese conclusion. Then render Japanese 変更点, 確認結果, and 次 paragraphs without bullets or decorative
-emoji. The plugin injects measured Speed, Cost, and 達成 paragraphs in a Japanese mission debrief card,
+emoji. The plugin injects measured Speed, Cost, and 達成 paragraphs in a collapsed Japanese mission debrief card,
 with one fixed icon per section, observed pack/model usage, validation/review, and evidence-backed traits.
 Its Markdown token bars and PACK RECORD summarize retained project goals, with coverage and team titles;
 they never imply lifetime history, XP, levels, unmeasured savings, or a leaderboard rank.
 Never write the card, its metrics, or its badges yourself. Do not estimate or fabricate them.
 Use 任務完了 for DONE, 中断帰還（未完了） for INTERRUPTED, 外部要因で待機（未完了） for BLOCKED,
 and 指示待ち（未完了） for NEED_DECISION; preserve the machine status token and first-line checkpoint.
-Never render a user-facing Evidence heading, <details> block, evidence reference, internal reason code,
+Never render a user-facing Evidence heading or Evidence details block, evidence reference, internal reason code,
 ledger key, or raw status. Keep ordered command/exit/fingerprint history, manifests, evidence refs,
 review proof, and terminal receipt append-only in their internal typed ledger and host logs. A concise
 確認結果 may summarize PASS/FAIL without exposing those internal identifiers.
 An undeclared write or mutation must be reported as rejected, not performed. A locally repairable process or evidence defect is never a
 user question: repair it and continue in the same turn.
+An INTERRUPTED root conclusion requires one machine line: \`TRUE_INTERRUPTION: user: <condition>\`
+for an explicit user stop, or \`TRUE_INTERRUPTION: internal: <condition>\` for a true internal
+interruption. Never emit that line for a local process defect, step boundary, continuation request, or
+recoverable limit; keep those on the same-session autonomous continuation path.
 
 TERMINAL_STATUS_SEMANTICS_FIXTURE
     DONE: all accepted criteria proved complete; unmet or interrupted work forbidden
@@ -1633,6 +1654,8 @@ TERMINAL_STATUS_SEMANTICS_FIXTURE
     status_icons: DONE=✅ | INTERRUPTED=⚠️ | BLOCKED=⛔ | NEED_DECISION=❓
     quality_gate_fail: validation evidence + autonomous non-adoption decision -> DONE; release remains unperformed
     process_defect: gate | routing | handoff | local tool defect -> autonomous repair; never terminal BLOCKED
+    interruption_marker: TRUE_INTERRUPTION: user: <condition> | TRUE_INTERRUPTION: internal: <condition>
+    continuation_not_interruption: local process defect | step boundary | continuation request | recoverable limit
 END_TERMINAL_STATUS_SEMANTICS_FIXTURE
 
 RUNTIME_ASSET_VERSION_SYNC_FIXTURE

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sortie-dogs",
-  "version": "0.9.7",
+  "version": "0.9.8",
   "description": "Bounded agent harness and validated orchestration loop plugin for OpenCode",
   "keywords": [
     "opencode",