harness-everything 0.21.0 → 0.22.0

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
       "name": "harness-everything",
       "source": "./",
       "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-      "version": "0.21.0",
+      "version": "0.22.0",
       "author": {
         "name": "Miya Daniel"
       },

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "harness-everything",
   "description": "Harness OS: an orchestrated agent operating system — tier routing, circuit breaker guardrails, verification gates, WAL session memory, and self-evolving workspace rules.",
-  "version": "0.21.0",
+  "version": "0.22.0",
   "author": {
     "name": "Miya Daniel"
   },

package/CHANGELOG.md

@@ -2,6 +2,18 @@
 
 All notable changes to this project are documented in this file.
 
+## [0.22.0](https://github.com/dyphn1/Harness-everything/compare/v0.21.0...v0.22.0) (2026-09-23)
+
+### Features
+
+* **router:** add resident system one provider ([eee8cb7](https://github.com/dyphn1/Harness-everything/commit/eee8cb763d2ed65722a2578c83700b02bdc3e668)), closes [#233](https://github.com/dyphn1/Harness-everything/issues/233)
+
+### Bug Fixes
+
+* **distribution:** keep python bytecode out of every package ([58135ae](https://github.com/dyphn1/Harness-everything/commit/58135aef92089d731eb44d55efb0ae77ea4d64dd)), closes [#233](https://github.com/dyphn1/Harness-everything/issues/233)
+* **router:** harden resident system one server from adversarial review ([ee5ab0f](https://github.com/dyphn1/Harness-everything/commit/ee5ab0fec9f03808f12a88e66b769b210803baa2)), closes [#233](https://github.com/dyphn1/Harness-everything/issues/233)
+* **router:** replace the whole windows dacl for resident state ([e59af73](https://github.com/dyphn1/Harness-everything/commit/e59af735de1337b3bcb10cbca05500ab6b03b64c)), closes [#233](https://github.com/dyphn1/Harness-everything/issues/233)
+
 ## [0.21.0](https://github.com/dyphn1/Harness-everything/compare/v0.20.7...v0.21.0) (2026-09-23)
 
 ### Features

package/benchmarks/results/system-one/cua-s1-forms-resident-cpu-2026-09-23/README.md

@@ -0,0 +1,74 @@
+# System One: resident CUA-S1-FORMS provider on CPU (2026-09-23)
+
+Evidence layer: **local real-checkpoint inference through the resident transport**.
+It is not live-host evidence, not a reviewed holdout, and not a Harness routing quality
+claim. Issue [#233](https://github.com/dyphn1/Harness-everything/issues/233); one-shot
+baseline: [cua-s1-forms-cpu-2026-09-23](../cua-s1-forms-cpu-2026-09-23/README.md).
+
+## Setup
+
+- Same host, venv and pinned checkpoint as the one-shot baseline: Intel Core Ultra 5 225H,
+  Windows 11 (win32 10.0.26200 x64), Node v24.18.0, CPython 3.13.6, torch `2.14.0+cpu`,
+  `cua-s1` at `b7f7e2d8…` (`pinned`), `cua-ai/cua-s1-forms` at `f54adbf4…`, one torch thread.
+- The installer was rerun with `transport: "resident"`: exit 0, `verification.ok: true`,
+  server ready in 7230 ms, first warm score 131 ms. The installer then stopped the server it had started.
+
+## Command
+
+```bash
+node scripts/evaluate-system-one.js benchmarks/fixtures/system-one-routing.json \
+  ~/.agents/harness-everything/system-one/manifest.json \
+  benchmarks/results/system-one/cua-s1-forms-resident-cpu-2026-09-23/report.json
+```
+
+[`report.json`](report.json) is the unmodified output (`evidence.kind: offline-resident`,
+`residentReady: true`, 16 holdout cases × 2 runs, 8 s wall versus 173 s one-shot).
+
+## Results
+
+| Metric | One-shot | Resident |
+| --- | --- | --- |
+| Samples | 32 cold | 32 warm |
+| p95 latency (end to end in Node) | 6415 ms | 108.6 ms |
+| p50 latency | — | 63.4 ms |
+| Repeatability (full vectors) | 1.0 | 1.0 |
+
+- Score vectors are byte-identical to the one-shot run for every case; decisions are
+  unchanged (all `domain-mismatch`, one 225-byte `provider-exit`), coverage 0,
+  `rolloutReady: false`.
+- The `warmLatency` gate (p95 ≤ 100 ms) is **not met** (108.6 ms).
+
+## Where the time goes (medians, same host)
+
+| Component | ms |
+| --- | --- |
+| Worker-thread bootstrap per call (sync bridge) | 29–41 |
+| Loopback connect | ≈3 |
+| Server `ping` / `score` measured from a Python client (back-to-back) | 1.2 / 7.5 |
+| Server `score` with 30 ms gaps between calls | 12.3 |
+| In-process warm inference (earlier probe) | 2.8–5.6 |
+
+Worker `env`/stdio options did not reduce bootstrap. Reusing the reply buffer lowered
+p50 (67 → 55 ms) but not p95, and was not kept.
+
+## Hook wall time
+
+`node harness-everything/scripts/kernel-router-core.js "fix a typo in README"`, 25
+interleaved runs per mode, with the server already warm:
+
+| Mode | p50 | p95 |
+| --- | --- | --- |
+| `off` | 273 ms | 409 ms |
+| `shadow` (resident) | 383 ms | 494 ms |
+
+Resident shadow mode adds roughly 85–110 ms per prompt. The one-shot bridge added 5–7 s.
+The router plan was identical to `off` in all 25 pairs. A cold start returns
+`provider-starting` without waiting.
+
+## Limitations
+
+The server survived the exit of the CLI processes that spawned it. Survival after
+exit of a real host's hook runner, and on macOS/Linux, is not verified. The seed
+corpus is 16 unreviewed cases. The forms checkpoint cannot drive Harness routing.
+Worker bootstrap is the remaining bottleneck. One possible follow-up is an async
+pre-score in the hook entry, which would remove the worker; this is not measured.