[Submission] lab-monitor@2.0.0

[x-dc-coder]

Author Studio manifest

{
  "schema": "omdsh-workshop-submission/v2",
  "operation": "create-project",
  "project": {
    "id": "lab-monitor",
    "repository": "https://github.com/x-dc-coder/lab-monitor",
    "path": null
  },
  "release": {
    "version": "2.0.0",
    "ref": "0edcf343c07a2f7830a4f92036c9d60979b625e5",
    "updatedAt": "2026-08-28T15:24:27.000Z",
    "channel": "beta",
    "compatibility": "Peer range ^0.1.0-rc.6; verified live on @deepseek-ai/dsh@0.1.1-rc.1; target baseline @deepseek-ai/dsh@0.1.0-rc.6",
    "changelog": "V2 formal plugin: dual-half TS build (host ESM + client CJS bundle), GPU/CPU/memory/proc sampling with Linux + native Windows backends, experiment state machine (multi-track <=4), balanced alert engine with graded advice, notify routing by experiment type, tag groups, settings.yaml persistence, conversation.view + better-sidebar + webServer data plane",
    "capabilities": {
      "requiresFabric": false,
      "deepHook": false,
      "restartRequired": true
    },
    "profileBundle": {
      "packageName": "lab-monitor",
      "spec": "github:x-dc-coder/lab-monitor#0edcf343c07a2f7830a4f92036c9d60979b625e5"
    },
    "updateFrom": null
  },
  "management": {
    "method": "profile-bundle",
    "protocol": "harness-profile",
    "label": "lab-monitor profile bundle (Host + Client two-half plugin)",
    "instructions": "Install the fixed source as a Profile Bundle. A web profile is required: webServer/settings/client-slot peers come from the @deepseek-ai/dsh-web-app bundle. Activation requires a profile restart (restart-profile). Removal deletes the dependency and reconciles the bundle list. See docs/WORKSHOP.md for the declared permission model and external side effects.",
    "source": null
  },
  "declarations": {
    "installScriptsMustRemainDisabled": true,
    "permissions": "process:observe (read-only GPU/CPU/memory/process sampling); shell:execute (built-in read-only sampler command sets: nvidia-smi/ps/tasklist/Windows CIM); tools:register (lab_status/lab_advice/lab_ctl, lab_ctl guarded to the monitoring engine only); webserver:register (localhost-only /lab-monitor/api/* routes); ui:extend (conversation.view tab + settings plugin card + optional better-sidebar tab, silent degradation)",
    "testing": "scripts/verify.sh: typecheck, build, directory/contract checks and seven regression suites (~190 assertions): verify-host.js (~140), mock-test.js, verify-m1.js (19), verify-overrides.js (10), verify-proc-detail.js (12), verify-backend-mode.js (6); optional real sampler and e2e runs gated by flags"
  },
  "packageManifest": {
    "schema": "omdsh-workshop-package/v1",
    "type": "plugin",
    "integration": {
      "protocol": "harness-profile",
      "artifact": "cordis.patch.yml"
    },
    "install": {
      "mode": "transactional",
      "adapter": "profile-bundle",
      "failurePolicy": "generation-rollback",
      "touchesCurrentBeforeActivation": false
    },
    "lifecycle": {
      "activation": "restart-profile",
      "dispose": "supported"
    },
    "permissions": [
      "process:observe",
      "shell:execute",
      "tools:register",
      "webserver:register",
      "ui:extend"
    ],
    "compatibility": {
      "dshVersions": [
        "0.1.1-rc.1"
      ]
    },
    "capability": {
      "id": "lab-status-snapshot",
      "kind": "tool",
      "invocation": "boot the RC.6 candidate Profile and call the lab_status tool with brief:true",
      "expected": "a JSON snapshot with ok:true containing GPU utilization, VRAM, CPU, memory and process entries"
    },
    "evidence": {
      "install": "docs/WORKSHOP.md",
      "failureIsolation": "docs/WORKSHOP.md",
      "hotReload": null,
      "remove": "docs/WORKSHOP.md"
    }
  }
}

Submission boundary

• This request contains only public, immutable source coordinates and the generated structured manifest.
• Automated intake may create a pending-review PR, but cannot approve the project or grant Registry installation authority.

Confirmations

• The repository and pinned commit are public.
• Permissions, tests, compatibility, and external effects are declared from verifiable evidence.
• No credential, private path, or private data is included.

[github-actions]

自动化插件预检未通过，尚未创建待审核记录。请查看 本次检查，修正固定来源或清单后重新提交。

[x-dc-coder]

Preflight failure root cause (author-side local replay)

The automated intake preflight failed on the repo-wide inventory invariant, not on this submission.

Passed: Issue manifest parsing, omdsh-workshop-submission/v2 schema, public fixed-source checks, and record preparation:

prepared lab-monitor@2.0.0 from Issue #91 with a typed Harness plan; Registry remains ineligible
intake accepted: 12 queued, 0 admitted, official baseline @deepseek-ai/dsh@0.1.0-rc.6

Failed: npm run validate → check-public-site.mjs:135 (verification inventory invariant).

Root cause (same self-block reported in #48 / #57 / #89, independently confirmed here on main 928cb55): line 127 expects untested === catalog.packages.length - intake.records.length, but build-verification-inventory.mjs can only classify records that map to a Catalog project and are already in a classified state. A freshly prepared pending-review record starts as untested and is not yet in the Catalog, so the two sides always differ by exactly 1 for any new submission. Local replay: catalog=680, records=12, classified=11 → inventory untested=669 vs expected=668 → FAIL. This is why recent submissions (#86–#90) all show the same error.

Fix proposed in #92 (fork PR, one-line formula change; mutation-tested: real desync still fails the gate): subtract only classified records. On today's main (11/11 classified) the formula is mathematically identical to the current one, so no behavior change for the existing tree; it only stops the false failure while a pending-review record is in flight.

This submission needs no change. Once the pipeline fix lands, re-running the intake workflow on this Issue should create the pending-review record/PR normally.