{
 "batch": "read1-111",
 "plugins": [
  {
   "pr": 0,
   "name": "done-bell",
   "does": "Plays an OS notification, chime and busy/done/needs-you status chip when a Hermes turn finishes or waits on approval.",
   "useful": 3,
   "reliable": 3,
   "polish": 3,
   "creative": 3,
   "beyond": 4,
   "total": 20,
   "best": "Uses session-end and approval hooks plus a Desktop pane and status chip, a small real quality-of-life gain no skill can provide.",
   "worst": "The README's install steps are Windows paths from the author's own machine plus a manual copy of the Desktop pane, and it has only 9 tests.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 0,
   "name": "homeassistant",
   "does": "Nous's official Home Assistant gateway platform adapter plus four tools that list, read and control smart-home devices.",
   "useful": 4,
   "reliable": 5,
   "polish": 4,
   "creative": 3,
   "beyond": 5,
   "total": 26,
   "best": "A full platform adapter (WebSocket event bus to the agent, notify delivery) with an entity denylist and extensive tests.",
   "worst": "Device control over a long-lived token is powerful, and safety depends on the user setting the denylist well.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 150594,
   "name": "write-gate",
   "does": "A pre_tool_call hook that blocks file writes with invalid JSON/TOML/Python, NUL bytes, oversized content or broken user regex contracts and returns a precise error to the model.",
   "useful": 4,
   "reliable": 4,
   "polish": 4,
   "creative": 3,
   "beyond": 4,
   "total": 23,
   "best": "Deterministic enforcement inside the tool loop that handles patch fragments carefully so they are not falsely rejected, with about 50 tests.",
   "worst": "Only a few syntaxes are checked, and it matches tool names and argument keys by guesswork, so it can miss write tools it does not recognise.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 153246,
   "name": "berrys",
   "does": "Points Hermes at Berrys' hosted OAuth MCP server for generating product listing images, ads and short videos, plus a long skill.",
   "useful": 3,
   "reliable": 2,
   "polish": 3,
   "creative": 2,
   "beyond": 3,
   "total": 16,
   "best": "A polished hosted creative service with OAuth sign-in and no keys to paste.",
   "worst": "There is no plugin code, only config files and a 20 KB skill for a remote paid service that does not appear at the pinned commit, and it is generic across many agents rather than made for Hermes.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 151884,
   "name": "fledly",
   "does": "Connects Hermes to Fledly's hosted MCP server for iOS App Store market data and keyword research, plus a routing skill.",
   "useful": 2,
   "reliable": 2,
   "polish": 3,
   "creative": 2,
   "beyond": 2,
   "total": 13,
   "best": "Clear multi-client install docs with OAuth through a pinned mcp-remote version.",
   "worst": "For Hermes it is only an mcp.json that runs npx (needing Node) and a skill, a niche data service with no Hermes-specific code or tests.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 152323,
   "name": "check-please",
   "does": "Matches each 'done' claim in a Hermes reply to the tool call that proves it and appends a backed/stale/not-backed line, with /check for detail and a Desktop pane.",
   "useful": 5,
   "reliable": 4,
   "polish": 4,
   "creative": 5,
   "beyond": 5,
   "total": 28,
   "best": "A new, deterministic claim-verification layer built on five hooks, with a big test suite, measured precision and recall, and candid limits.",
   "worst": "Screenshots are placeholders and recall (0.768) is admittedly short, and a 35 KB __init__.py is heavy to review.",
   "standout": true,
   "blocked": null
  },
  {
   "pr": 153964,
   "name": "rule-injector",
   "does": "Injects user-defined trigger-matched rules at pre_llm_call, into tool results, and at the pre_verify gate instead of keeping them in the system prompt.",
   "useful": 3,
   "reliable": 3,
   "polish": 3,
   "creative": 4,
   "beyond": 4,
   "total": 21,
   "best": "Puts rules at the point of need, with compound triggers and a demo video and dashboard.",
   "worst": "The repo is cluttered with one-off stage migration scripts, the manifest has no author, and many top-level test files hold no test functions.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 150603,
   "name": "getyoutubetranscript",
   "does": "Connects Hermes to the GetYouTubeTranscript remote MCP server for transcripts, YouTube search and channel browsing, plus four skills.",
   "useful": 3,
   "reliable": 3,
   "polish": 3,
   "creative": 2,
   "beyond": 2,
   "total": 15,
   "best": "Honest disclosure of what is sent, the credit cost, and treating transcripts as data.",
   "worst": "It is only configuration and skills, with the API key read from a shell or .env variable, and the tool set is served remotely and not covered by the pinned commit.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 152751,
   "name": "groundcontrol",
   "does": "Adds a /groundcontrol command and an operating skill for using GroundControl's scoped MCP deployment tools, with a conservative evidence-receipt reviewer.",
   "useful": 2,
   "reliable": 3,
   "polish": 3,
   "creative": 3,
   "beyond": 2,
   "total": 15,
   "best": "Careful safety stance: read-only by default, OAuth stays in Hermes, and receipts never echo secrets.",
   "worst": "The code is a short command that returns guidance text and a JSON projection, so it adds almost nothing a skill could not do, and it needs a GroundControl server.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 152892,
   "name": "autom",
   "does": "Adds Autom's Google SERP API as a Hermes web_search backend provider.",
   "useful": 3,
   "reliable": 4,
   "polish": 3,
   "creative": 2,
   "beyond": 4,
   "total": 20,
   "best": "A clean native web-search provider: the key comes from Hermes's secret store, it uses only the standard library, and its tests mock the network.",
   "worst": "One more paid search backend, with search only (no extract).",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 151199,
   "name": "hermes-australian-legal",
   "does": "Five tools for dated Australian regulatory provision lookup, review gates, exact-decimal earn-out arithmetic and quote checks, plus a counsel-preparation skill.",
   "useful": 2,
   "reliable": 3,
   "polish": 4,
   "creative": 4,
   "beyond": 3,
   "total": 19,
   "best": "Evidence-linked, dated provision reads that refuse dates past the corpus, with tests and careful scope docs.",
   "worst": "A self-described experimental prototype for a narrow legal niche that depends on an external Git corpus and says it is not for real use.",
   "standout": false,
   "blocked": null
  },
  {
   "pr": 153986,
   "name": "engagements",
   "does": "A file-based client and project memory in Markdown: tools to find, bind, create and update clients, a pre_llm_call banner, and binding of sessions by working directory.",
   "useful": 4,
   "reliable": 4,
   "polish": 4,
   "creative": 4,
   "beyond": 4,
   "total": 24,
   "best": "A thoughtful, heavily tested memory layer that binds sessions to clients automatically and keeps records in human-editable Markdown.",
   "worst": "An rc1 release with complex binding rules, and the repo ships a spike/ probe plugin alongside the real one.",
   "standout": false,
   "blocked": null
  }
 ]
}