{
  "batch": "deep-58",
  "setup": "Cloned Hermes source into test/scratch/hermes and ran uv sync --locked with a scratch-local Python 3.14; hermes plugins --help exposed install, validate and doctor. Direct catalog validation passed for every entry. The supplied tryit.py was attempted, but activation bootstraps a separate per-profile Hermes runtime and dependency environment; that bootstrap exceeded this executor's 30-second per-process limit before a report or scripted chat could complete. Scores therefore use the successful catalog capability probes, code inspection, and the stated test runs; no unobserved tool call is claimed.",
  "plugins": [
    {
      "pr": 149685,
      "name": "hermes-teleoperator",
      "validate": "passed: manifest, loadability, isolated capability probe, declared registrations, collision, security, core-override, and desktop checks passed (with a private-core-import warning).",
      "inside": "Not loaded into a completed Hermes chat: tryit activation did not finish its isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe did load register() in isolation. No planned calls were observed.",
      "tests": "46 passed, 40 subtests passed (targeted plugin/controller/guard suite).",
      "useful": 4,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "total": 28,
      "rank": 1,
      "best": "It safely makes one Hermes conversation operate across configured SSH machines, including their native tools and files.",
      "worst": "Its value depends on an SSH-accessible Unix machine, and it retains private Hermes approval imports flagged by validation.",
      "blocked": null
    },
    {
      "pr": 152258,
      "name": "kumi",
      "validate": "passed: manifest, schema, loadability, isolated capability probe, declared registrations, collision, security, core-override, and desktop checks passed (with one private-core-import warning).",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "96 passed and 238 subtests passed; 1 failed because its spawned Hermes-approval check imported the plugin's top-level tools.py instead of Hermes's tools module in this test layout.",
      "useful": 4,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "total": 28,
      "rank": 2,
      "best": "It turns Hermes into a deeply integrated, approval-aware and undoable assistant for three major music-production environments.",
      "worst": "Its substantial capability is useful chiefly to DAW users, and the test layout exposes a top-level module-name collision.",
      "blocked": null
    },
    {
      "pr": 153960,
      "name": "on-air",
      "validate": "passed: manifest, schema, loadability, isolated capability probe, declared registrations, collision, security, core-override, and desktop checks all passed.",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "19 passed (targeted manifest-warning and runtime-adapter tests); the broader selected suite did not finish before the executor limit.",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "total": 27,
      "rank": 4,
      "best": "It is an unusually complete live-show platform, combining guarded co-hosting, Twitch reading, OBS output, and an approval desk.",
      "worst": "Live-streaming and OBS setup make it a specialized install rather than an everyday Hermes default.",
      "blocked": null
    },
    {
      "pr": 153240,
      "name": "flowcanvas",
      "validate": "passed: manifest, schema, loadability, isolated capability probe, declared registrations, collision, security, and core-override checks passed; validation warned about raw window.hermesDesktop use.",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "564 passed, 2 failed, 4 errored, and 2 skipped (572 total); the errors/failures shown used AF_UNIX multiprocessing sockets blocked by this sandbox.",
      "useful": 4,
      "reliable": 3,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "rank": 5,
      "best": "Its finished Desktop canvas makes reusable multi-tool and model workflows inspectable and runnable rather than merely prompt-driven.",
      "worst": "The large test suite did not pass cleanly here and catalog validation also flagged use of a raw Desktop bridge.",
      "blocked": null
    },
    {
      "pr": 150453,
      "name": "workgraph",
      "validate": "passed: manifest, schema, loadability, isolated capability probe, declared registrations, collision, security, and core-override checks all passed.",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "7 passed and 8 failed in targeted engine/interface tests; one interface test received 502 rather than 401 under the sandbox proxy, while several candidate-flow tests also failed.",
      "useful": 3,
      "reliable": 2,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "total": 25,
      "rank": 6,
      "best": "It gives agents a distinctive persistent product-state graph for branching, comparing, combining, and replaying development work.",
      "worst": "Its core targeted tests had several failures, leaving the versioning workflow less trustworthy than its polished presentation suggests.",
      "blocked": null
    },
    {
      "pr": 149611,
      "name": "hermes-security",
      "validate": "passed: manifest, loadability, isolated capability probe, declared registrations, collision, security, core-override, and desktop checks all passed.",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "363 passed and 1 failed; the failure was a process-writer test blocked when the sandbox refused an AF_UNIX socket.",
      "useful": 4,
      "reliable": 4,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "total": 27,
      "rank": 3,
      "best": "It adds a substantial local, evidence-oriented security-review workflow with controlled validation and a dedicated Desktop surface.",
      "worst": "It still relies on a model-led review process and its process-backed storage path could not be exercised in this sandbox.",
      "blocked": null
    },
    {
      "pr": 0,
      "name": "done-bell",
      "validate": "passed: manifest, schema, loadability, isolated capability probe, declared registrations, collision, security, core-override, and desktop checks all passed.",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "9 passed.",
      "useful": 3,
      "reliable": 5,
      "polish": 4,
      "creative": 2,
      "beyond": 3,
      "total": 20,
      "rank": 0,
      "best": "It is a small, focused notification and status plugin with a clean tested implementation.",
      "worst": "Turn-completion alerts are helpful but do not materially expand what an agent can accomplish.",
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "validate": "passed: manifest, loadability, isolated capability probe, declared registrations, trusted-inbound platform, collision, security, and core-override checks passed (with a private-core-import warning).",
      "inside": "Not loaded into a completed Hermes chat: the shared tryit activation path could not finish isolated-runtime bootstrap in the 30-second executor window; the catalog capability probe loaded register() in isolation. No planned calls were observed.",
      "tests": "49 passed, 1 skipped (targeted registration and service-policy tests).",
      "useful": 4,
      "reliable": 5,
      "polish": 5,
      "creative": 3,
      "beyond": 5,
      "total": 27,
      "rank": 0,
      "best": "It is a mature platform integration that lets Hermes safely observe and control a user's own Home Assistant devices.",
      "worst": "It requires an existing Home Assistant deployment and token, so its audience is bounded by that ecosystem.",
      "blocked": null
    }
  ]
}