{
  "batch": "read1-098",
  "plugins": [
    {
      "pr": 0,
      "name": "done-bell",
      "does": "Signals when a Hermes turn finishes or needs approval through OS notifications, a chime, and a Desktop status pane.",
      "useful": 3,
      "reliable": 3,
      "polish": 3,
      "creative": 2,
      "beyond": 4,
      "total": 19,
      "best": "Hooks, status tracking, mute controls, OS alerts, and a Desktop pane make the signal visible beyond an in-chat prompt.",
      "worst": "The README's Desktop setup includes manually copying the pane file, and its small test suite does not exercise the Desktop UI.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "does": "Connects Hermes to Home Assistant events and exposes tools to inspect entities, read state, list services, and control devices.",
      "useful": 5,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 27,
      "best": "It turns Hermes into an event-aware smart-home controller, with entity scoping, actionable service errors, and extensive focused tests.",
      "worst": "It requires a Home Assistant instance and long-lived token, and exposed service calls can control devices without an additional plugin-level human confirmation step.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150446,
      "name": "global-business-culture",
      "does": "Provides country business-culture briefs and task-weighted home-versus-host gap analysis from bundled framework data.",
      "useful": 3,
      "reliable": 3,
      "polish": 4,
      "creative": 3,
      "beyond": 3,
      "total": 19,
      "best": "The tools compute and rank cross-cultural differences locally, with country lookup, task weighting, input handling, and tests covering data and calculations.",
      "worst": "Country scores and advice are author-encoded approximations with limited country-level source traceability, so confident-looking comparisons may overstate their evidence.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152383,
      "name": "draftroom",
      "does": "Adds a Hermes Desktop Markdown workspace where the agent proposes selected-text rewrites and the editor user accepts or rejects them.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "best": "Version-bound proposals, optimistic saves, atomic writes, revision history, native Desktop controls, and extensive backend and controller tests support a real co-writing loop.",
      "worst": "It is an early release with known UX and accessibility work, and the backend requires Linux or macOS plus a compatible recent Desktop SDK.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 152876,
      "name": "oneiros",
      "does": "Reviews session and memory facts into proposals for Hermes memory files, then applies accepted changes with snapshots, receipts, and rollback.",
      "useful": 5,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 27,
      "best": "Memory changes remain reviewable and gated, with evidence checks, secret scans, locking, atomic writes, snapshots, rollback, and a broad regression suite.",
      "worst": "The feature is complex and depends on host hooks and session-log integration; the README says the Dashboard has no pinned screenshot URLs, and older hosts may warn on an unknown hook.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 151992,
      "name": "kyte",
      "does": "Packages a remote Kyte MCP options market and two skills for market reads, trade preparation, and scheduled alerts for actions awaiting the user's signature.",
      "useful": 2,
      "reliable": 3,
      "polish": 4,
      "creative": 3,
      "beyond": 3,
      "total": 18,
      "best": "It separates read-only market queries from prepare-action links, labels the MCP server untrusted, and keeps signing in the user's own wallet.",
      "worst": "Live tool behavior is supplied by kyte.sh rather than the pinned repository, and the feature depends on an external finance service with no local tests or implementation to inspect.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152871,
      "name": "hermes-maps",
      "does": "Pairs nine Hermes tools with a Desktop map where the agent plots geocoded places and routes and the user can send point or area selections back to chat.",
      "useful": 4,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "total": 28,
      "best": "The map is a genuinely shared agent surface with interactive user-to-agent selections, persistent scenes, bounded tool waits, place-match safeguards, screenshots, and extensive tests.",
      "worst": "Place names, queries, map bounds, and tile requests leave the machine for third-party services, and those services have no uptime guarantee.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 152966,
      "name": "board-insights",
      "does": "Adds a Desktop List and Dashboard for current Kanban work and historical flow, trends, run times, and assignee workload across local boards.",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 3,
      "beyond": 5,
      "total": 25,
      "best": "A polished, screenshot-documented Desktop product turns Hermes's own Kanban data into useful live and historical views while opening every board database read-only.",
      "worst": "It serves Hermes Kanban users only, has no agent tools or model interaction, and requires separately enabling both the plugin and Desktop page.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 150663,
      "name": "fast-chat",
      "does": "Adds a CLI command that creates a fresh Hermes session with a chosen model, provider, reasoning level, and Fast or Ultrafast service tier.",
      "useful": 3,
      "reliable": 4,
      "polish": 3,
      "creative": 3,
      "beyond": 5,
      "total": 23,
      "best": "It checks the Gateway session contract and tier availability before opening a frontend, with bounded RPC/bootstrap waits and real local WebSocket tests.",
      "worst": "It depends on recent Hermes Gateway APIs and websockets packaging, is verified live only on macOS, and may leave an empty session if tier validation fails.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152053,
      "name": "compound-security",
      "does": "Provides 13 security workflow skills with local scripts for dependency evidence, audit records, patch tracking, verification, and reusable prevention.",
      "useful": 4,
      "reliable": 3,
      "polish": 4,
      "creative": 4,
      "beyond": 3,
      "total": 21,
      "best": "It combines scoped audit and remediation workflows with tested local validators and explicit evidence records, and its docs state the limits of each stage.",
      "worst": "It supplies no sandbox or scanners, has no end-to-end host/model audit validation, and most of its behavior is skill guidance that relies on the host's shell and agent capabilities.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152361,
      "name": "leadpoet",
      "does": "Turns a seller's website into a sourced prospect brief, personalized outreach drafts, person-approved campaign launches, and reply drafting in chat.",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "best": "The workflow pairs evidence checks for personalized claims with cost disclosure, secret-store API access, code-bound human approval, and 213 focused tests.",
      "worst": "The current PR text identifies v0.2.7 at a different SHA than this batch's v0.2.6 pin, so the reviewed tree is older than the PR description says.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 152979,
      "name": "lydite",
      "does": "Hands an approved coding task and proof command to sandboxed external workers, replays the proof, and opens a reviewed pull request with a receipt.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 5,
      "beyond": 5,
      "total": 27,
      "best": "A human approval hook exposes the repository, task, proof, and target before launch; tools keep secrets out of logs and use a closed child environment, while tests exercise refusals and plan fingerprints.",
      "worst": "This repository is only the Hermes shim: users must separately install and configure the attested Lydite binary, Claude Code, GitHub CLI, and sandbox prerequisites; the integration is Linux-only.",
      "standout": true,
      "blocked": null
    }
  ]
}