@hmharness/agent 0.22.3 → 0.22.4

package/dist/runner.js

@@ -133,7 +133,14 @@ export async function contextPack(task, sessionId, opts = {}) {
         return { memory: '', skills: '', insights: '', cognitive: '', skillsInjected: [] };
     }
     const [memory, skills, insights, cognitive] = await Promise.all([
-        retrieveMemory(home, task, { workspace: opts.workspace ?? undefined, embedding: opts.embedding }),
+        // HMH_NO_CONTEXT=1 strips the harness EXPERIENCE layer (memory +
+        // task-relevant insights) — the full-ablation arm for §26. The v0.22.3
+        // reading showed tool-reliability digest alone has no result effect;
+        // the live hypothesis is that the VALUE lives in the lessons layer,
+        // which was present in BOTH arms of every ablation so far.
+        process.env.HMH_NO_CONTEXT === '1'
+            ? Promise.resolve('')
+            : retrieveMemory(home, task, { workspace: opts.workspace ?? undefined, embedding: opts.embedding }),
         listSkills(home),
         // 2026-09-27 plan A: experience retrieval - the K most RELEVANT past
         // insights for THIS task (bigram similarity, recency tiebreak) instead
@@ -141,7 +148,9 @@ export async function contextPack(task, sessionId, opts = {}) {
         // surfaces that lesson instead of five unrelated recent rows.
         // Plan C: cross-project lessons stay retrievable but carry their
         // origin stamp ([来自项目:X]) so the model can weigh them.
-        retrieveInsights(home, task, { workspace: opts.workspace ?? undefined }),
+        process.env.HMH_NO_CONTEXT === '1'
+            ? Promise.resolve('')
+            : retrieveInsights(home, task, { workspace: opts.workspace ?? undefined }),
         // blueprint M3: the world model feeds the planner — which tools history
         // trusts/distrusts lands in the system prompt (empty until ≥3 steps of
         // evidence exist, so cold starts see nothing). HMH_NO_COGNITIVE=1 strips

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hmharness/agent",
-  "version": "0.22.3",
+  "version": "0.22.4",
   "description": "hmharness agent execution layer: base tools, system prompt, sub-agent spawn, and the shared task runner that frontends (cli, web) drive.",
   "type": "module",
   "main": "dist/index.js",
@@ -15,13 +15,13 @@
     "build": "tsc -p tsconfig.build.json"
   },
   "dependencies": {
-    "@hmharness/cognitive": "0.22.3",
-    "@hmharness/domain-harmony": "0.22.3",
-    "@hmharness/domain-ops": "0.22.3",
-    "@hmharness/evolution": "0.22.3",
-    "@hmharness/kernel": "0.22.3",
-    "@hmharness/observability": "0.22.3",
-    "@hmharness/sandbox": "0.22.3"
+    "@hmharness/cognitive": "0.22.4",
+    "@hmharness/domain-harmony": "0.22.4",
+    "@hmharness/domain-ops": "0.22.4",
+    "@hmharness/evolution": "0.22.4",
+    "@hmharness/kernel": "0.22.4",
+    "@hmharness/observability": "0.22.4",
+    "@hmharness/sandbox": "0.22.4"
   },
   "files": [
     "dist"