sortie-dogs 0.14.6 → 0.14.8

package/README.md

@@ -19,10 +19,10 @@ evidence-backed result, not more agents for their own sake.
 [![Node.js](https://img.shields.io/node/v/sortie-dogs)](https://www.npmjs.com/package/sortie-dogs)
 [![MIT License](https://img.shields.io/npm/l/sortie-dogs)](LICENSE)
 
-**Current release: [v0.14.6](https://github.com/zufall-upon/Sortie-dogs/releases/tag/v0.14.6)**
-· [Release notes](docs/release-v0.14.6.md) · [Execution vision](docs/v0.14.0-execution-vision.md)
+**Current release: [v0.14.8](https://github.com/zufall-upon/Sortie-dogs/releases/tag/v0.14.8)**
+· [Release notes](docs/release-v0.14.8.md) · [Execution vision](docs/v0.14.0-execution-vision.md)
 
-[Quick start](#quick-start) · [Workflow](#mission-workflow) · [Models](#default-models) ·
+[When to use](#when-to-use-sortie-dogs) · [Quick start](#quick-start) · [Workflow](#mission-workflow) · [Models](#default-models) ·
 [Measured results](#measured-results) · [Configuration](#configuration) · [Documentation](#documentation)
 
 > **Beta:** the package is still stabilizing before 1.0. Both host integrations are
@@ -46,6 +46,26 @@ evidence-backed result, not more agents for their own sake.
 - **Coexistence:** explicit invocation, existing authentication and host settings, one shared
   Mission lifecycle. Normal OpenCode/Codex use and user-owned files remain yours.
 
+## When to use Sortie-dogs
+
+Use Sortie with an existing **OpenCode V2 or Codex CLI** when a bounded coding task needs:
+
+- The original goal and constraints retained across investigation, multi-file implementation,
+  validation, review and correction.
+- Either a direct author or delegated work, with edits, valid checks and cumulative budget retained
+  when the author changes.
+- A handoff that distinguishes actual commands, test results, unresolved issues and acceptance
+  from a model's completion claim.
+
+For example, request a bug fix with a regression test, preservation of unrelated local work, and
+delivery as a reviewed PR. Define the required checks and delivery boundary in the request.
+
+**Use the host directly** for ordinary questions or small edits that do not need this lifecycle.
+Sortie is not a replacement for either host, a way to bypass native permissions, or a guarantee of
+unattended completion, lower cost or better benchmark scores. Budget, model access and environment
+requirements still matter. See [host setup](#quick-start) and [evidence and limitations](#measured-results)
+before choosing it.
+
 ## Quick start
 
 Choose **one host route**. Both require Node.js **22.6+** and npm; project-local installation is
@@ -57,7 +77,7 @@ Use an existing Codex CLI signed in with **ChatGPT**. On Windows, PowerShell **7
 available; POSIX uses bash. Sortie does not install Codex, start login or use a metered API fallback.
 
 ```sh
-npm install --save-dev sortie-dogs@0.14.6
+npm install --save-dev sortie-dogs@0.14.8
 npx --no-install sortie-dogs codex init .
 ```
 
@@ -86,7 +106,7 @@ bridge. [Codex guide: SDK, permissions, progress, resume and manifest route](doc
 Use **OpenCode V2**. In the target project:
 
 ```sh
-npm install --save-dev sortie-dogs@0.14.6
+npm install --save-dev sortie-dogs@0.14.8
 npx --no-install sortie-dogs init .
 ```
 
@@ -159,6 +179,12 @@ inferred from token totals.
 
 ## Measured results
 
+**Evidence and limitations:** these are recorded runs with fixed packages, models, tasks and
+execution conditions, not a general performance guarantee or a controlled comparison with the
+native hosts or other harnesses. Historical results do not measure the current release or its
+current default Worker. Mission acceptance, official task scoring and leaderboard acceptance are
+different outcomes; the records below keep them separate.
+
 ### Codex end-to-end completion
 
 **Codex integration has completed real Missions on Ubuntu and Windows.** The Windows large-write
@@ -245,7 +271,7 @@ Install the desired package version and rerun **your host's** initializer:
 
 An OpenCode config-local bridge may resolve a separate package; npm-global update alone does not
 update it. [Global/update instructions](docs/configuration.md#global-availability).
-The current Mission asset marker is `0.14.6-attachment-terra-v1`; the Codex skill marker is
+The current Mission asset marker is `0.14.8-discovery-fit-v1`; the Codex skill marker is
 `0.14.6-codex-skill-v1`. Installed markers alone do not prove a running host reloaded the new version.
 
 There is no supported uninstall command. Remove the dependency and only known Sortie-owned paths;
@@ -261,9 +287,14 @@ follow the [manual removal guide](docs/uninstall.md), never delete the entire `.
   [SWE-bench history](docs/benchmarks/swebench-lite-history.md) · [Historical local case study](docs/benchmark-reference.md).
 - **Development:** [Testing](docs/testing.md) · [CLI testing](docs/cli-testing.md) ·
   [Windows tests](docs/windows-tests.md) · [Release routine](docs/release-batch.md).
-- **Releases:** [v0.14.6](docs/release-v0.14.6.md) · [v0.14.5](docs/release-v0.14.5.md) · [v0.14.4](docs/release-v0.14.4.md) · [v0.14.3](docs/release-v0.14.3.md) · [v0.14.2](docs/release-v0.14.2.md) · [v0.14.1](docs/release-v0.14.1.md) · [v0.14.0](docs/release-v0.14.0.md) · [v0.13.22](docs/release-v0.13.22.md) · [v0.13.21](docs/release-v0.13.21.md) · [v0.13.20](docs/release-v0.13.20.md) · [v0.13.19](docs/release-v0.13.19.md) · [v0.13.18](docs/release-v0.13.18.md) · [v0.13.17](docs/release-v0.13.17.md) · [v0.13.16](docs/release-v0.13.16.md) · [v0.13.15](docs/release-v0.13.15.md) · [v0.13.14](docs/release-v0.13.14.md) · [v0.13.13](docs/release-v0.13.13.md) · [v0.13.12](docs/release-v0.13.12.md) · [v0.13.11](docs/release-v0.13.11.md) · [v0.13.10](docs/release-v0.13.10.md) ·
+- **Releases:** [v0.14.8](docs/release-v0.14.8.md) · [v0.14.7](docs/release-v0.14.7.md) · [v0.14.6](docs/release-v0.14.6.md) · [v0.14.5](docs/release-v0.14.5.md) · [v0.14.4](docs/release-v0.14.4.md) · [v0.14.3](docs/release-v0.14.3.md) · [v0.14.2](docs/release-v0.14.2.md) · [v0.14.1](docs/release-v0.14.1.md) · [v0.14.0](docs/release-v0.14.0.md) · [v0.13.22](docs/release-v0.13.22.md) · [v0.13.21](docs/release-v0.13.21.md) · [v0.13.20](docs/release-v0.13.20.md) · [v0.13.19](docs/release-v0.13.19.md) · [v0.13.18](docs/release-v0.13.18.md) · [v0.13.17](docs/release-v0.13.17.md) · [v0.13.16](docs/release-v0.13.16.md) · [v0.13.15](docs/release-v0.13.15.md) · [v0.13.14](docs/release-v0.13.14.md) · [v0.13.13](docs/release-v0.13.13.md) · [v0.13.12](docs/release-v0.13.12.md) · [v0.13.11](docs/release-v0.13.11.md) · [v0.13.10](docs/release-v0.13.10.md) ·
   [v0.13.9](docs/release-v0.13.9.md) · [All GitHub releases](https://github.com/zufall-upon/Sortie-dogs/releases).
 
+## Support
+
+[GitHub Sponsors](https://github.com/sponsors/zufall-upon) — sponsorships become available
+once the maintainer's Sponsors profile is approved and published.
+
 ## Community
 
 [Contributing](CONTRIBUTING.md) · [Code of conduct](CODE_OF_CONDUCT.md) ·

package/dist/asset-version.d.ts

@@ -3,6 +3,6 @@
  * installed project marker without importing every asset body.
  */
 export declare const RUNTIME_ASSET_VERSION = "0.3.89-completion-proof-v1";
-export declare const V010_RUNTIME_ASSET_VERSION = "0.14.6-attachment-terra-v1";
+export declare const V010_RUNTIME_ASSET_VERSION = "0.14.8-discovery-fit-v1";
 export declare const CODEX_SKILL_ASSET_VERSION = "0.14.6-codex-skill-v1";
 export type RuntimeAssetVersion = typeof RUNTIME_ASSET_VERSION | typeof V010_RUNTIME_ASSET_VERSION | typeof CODEX_SKILL_ASSET_VERSION;

package/dist/asset-version.js

@@ -3,5 +3,5 @@
  * installed project marker without importing every asset body.
  */
 export const RUNTIME_ASSET_VERSION = "0.3.89-completion-proof-v1";
-export const V010_RUNTIME_ASSET_VERSION = "0.14.6-attachment-terra-v1";
+export const V010_RUNTIME_ASSET_VERSION = "0.14.8-discovery-fit-v1";
 export const CODEX_SKILL_ASSET_VERSION = "0.14.6-codex-skill-v1";

package/dist/plugin/index.js

@@ -8488,12 +8488,16 @@ export const SortieDogsPlugin = async (input, options) => {
             return settled;
         },
         releaseDirectUnit: async (actor) => {
-            await releaseWriteGate(actor);
+            const released = JSON.parse(await releaseWriteGate(actor));
+            if (released.status === "denied")
+                throw new Error(`mission-direct-release:${released.reason}`);
             releaseSessionEnforcement(actor);
         },
         assertDirectUnitQuiescent: async (actor) => {
             if (activeSessions.get(actor)?.inFlightCalls.size)
                 throw new Error("mission-direct-tools-running");
+            if (bindingOperations.has(actor))
+                throw new Error("mission-direct-binding-in-flight");
         },
         failDirectUnit: async (root, actor) => {
             const direct = await input.runtimeBridge?.directMissionExecution?.(actor);

package/dist/plugin/profiled.js

@@ -1493,8 +1493,8 @@ export function createProfiledPlugin(profile, assetVersion) {
         }
         async function stop(root, reason, retireRoot = true) {
             const directActor = (await operators.read(root))?.units.find(unit => unit.status === "running" && unit.directExecution)?.directExecution?.actor;
-            if (directActor)
-                await activateDirect(directActor);
+            // Stopping owned execution does not write its artifacts. Its existing reservation
+            // is reconciled by cancelChildren, even when activation never acquired a writer.
             if (retireRoot)
                 retired.add(root);
             if (retireRoot)
@@ -2290,7 +2290,9 @@ export function createProfiledPlugin(profile, assetVersion) {
                     record(part.state.metadata) && ["running", "pending", "unknown"].includes(String(part.state.metadata.status)))) {
                     return JSON.stringify({ status: "direct-handoff-awaits-tools", next_action: "Finish the existing native command before handing off this unit; retain its result and continue here or delegate afterwards." });
                 }
-                await activateDirect(actor);
+                // Yielding execution is management, not an artifact write. In particular, a
+                // direct admission can exist even when its writer could not bind because another
+                // live owner overlaps. Do not require that contested binding to hand it off.
                 const mission = await missions.retainContext(root, actor, history, notes);
                 await operators.handoffDirect(root, actor, await missions.handoffContext(mission));
             }
@@ -3600,7 +3602,7 @@ export function createProfiledPlugin(profile, assetVersion) {
                     else if (args.status === "blocked" || args.status === "needs-decision") {
                         // This owner is stopping its direct work, not validating it. Keep the
                         // failed checks/candidate and settle its existing reservation once.
-                        const direct = await activateDirect(context.sessionID);
+                        const direct = await directExecution(context.sessionID);
                         if (direct)
                             await control.failDirectUnit(root, context.sessionID);
                     }
@@ -4015,7 +4017,11 @@ export function createProfiledPlugin(profile, assetVersion) {
                     throw new Error(RUNTIME_PROFILE_SESSION_INACTIVE);
                 if (request.tool.startsWith("sortie_") && !request.tool.startsWith(profile.toolPrefix))
                     throw new Error("runtime-profile-tool-mismatch");
-                const direct = await activateDirect(request.sessionID);
+                // Management can reconcile, yield, stop or report the owner's admission.
+                // They must remain reachable when its artifact binding is the thing in conflict.
+                // Ordinary tools and the parts of planning that actually write still activate it.
+                const direct = request.tool === status || request.tool === planUnits || request.tool === cancel || request.tool === submitMission
+                    ? await directExecution(request.sessionID) : await activateDirect(request.sessionID);
                 const writer = who.role === "dog-worker" || !!direct || !!await reviewerCorrection(request.sessionID);
                 if (who.role === "dog-reviewer" && writer) {
                     const correction = (await reviewerCorrection(request.sessionID));

package/dist/runtime-mission-assets.js

@@ -155,6 +155,17 @@ ${MISSION_DEPENDENCY_WORKFLOW}
     SAME mission before any Worker starts, retains its Coordinator and cumulative spend, and checks
     old children before preparing a Worker. This is not permission to discard unchanged acceptance:
     if the scope is uncertain, ask the user which requirements remain instead of inferring a replacement.
+    When execution management is inconsistent or manifest-overlap blocks the owning direct writer,
+    explain the observed stop, what can be reconciled, and that the existing requirements, results,
+    checks and cumulative spend remain. Offer to recover and continue in ordinary conversation;
+    an ordinary affirmative reply or a request to self-recover is enough. No hidden command or ID is
+    needed from the user, and ordinary autonomous recovery does not require a new approval.
+    Inspect operator_status to reconcile only native-proven ended work. If this session owns a quiescent
+    direct admission that cannot bind, plan_units with executor: "worker" and no units yields that same
+    unit and reservation without acquiring the contested writer. Explain the result and continue the
+    returned Task when its write scope is available. A genuinely live or unproven other owner is not
+    released: wait or separate work, and do not dispatch repeated conflicting attempts. Do not invent
+    terminal state, validation success, a fresh Mission, or a budget reset.
     If a Coordinator reports mission-replan-worker-still-active after an interrupted native Task,
     it cannot call cancel_operator (root-only). Inspect operator_status and the native Task outcome.
     If the Worker is still active, wait, not duplicate. If the user chose to stop/replan, root calls
@@ -494,6 +505,10 @@ in this SAME session when your existing context makes that efficient. You are no
 Use the Operator's preferred_executor as a starting recommendation. Start useful work with available context;
 do not investigate merely to confirm the recommendation. Adapt through actual work, honoring explicit user choices.
 Use plan_units with executor: "self" for direct work, or start_direct_unit for an already prepared unit.
+If a direct admission cannot bind because of manifest-overlap, operator_status and plan_units with
+executor: "worker" and no units remain management operations: yield only this quiescent owner, retain
+its checks and reservation, and wait for a real conflicting writer rather than evicting it. Continue
+the same Mission and actual remaining work; do not create another Mission to clear the conflict.
 The host binds your native check results to the candidate; finish_direct_unit closes that work without
 a Worker handoff. Independent Review and Operator final acceptance still follow. This is not limited to
 micro-edits. Delegate to the configured Worker when that reduces work/cost or provides useful separation.

package/package.json

@@ -1,10 +1,12 @@
 {
   "name": "sortie-dogs",
-  "version": "0.14.6",
-  "description": "Bounded agent harness and validated orchestration loop plugin for OpenCode",
+  "version": "0.14.8",
+  "description": "Adaptive execution harness for OpenCode and Codex: goal retention, validation, review, and evidence-backed handoff",
   "keywords": [
     "opencode",
     "opencode-plugin",
+    "codex",
+    "codex-cli",
     "ai-orchestration",
     "multi-agent",
     "agent-harness",