{
  "batch": "read1-080",
  "plugins": [
    {
      "pr": 0,
      "name": "done-bell",
      "does": "Notifies users when Hermes finishes a turn or needs approval, with a chime, Desktop status chip and recent-event history.",
      "useful": 4,
      "reliable": 3,
      "polish": 3,
      "creative": 2,
      "beyond": 4,
      "total": 20,
      "best": "Lifecycle hooks and a Desktop watcher let users leave a running agent without repeatedly checking it.",
      "worst": "Setup includes a developer-specific Windows path and manual Desktop copying, while tests do not exercise the Desktop notification path.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "does": "Connects Hermes to Home Assistant device queries, service calls, state-change events and notification delivery.",
      "useful": 4,
      "reliable": 5,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "total": 26,
      "best": "A real gateway platform with scoped secrets, reconnect handling and tested entity/service restrictions makes Hermes useful around the home.",
      "worst": "Its value requires a Home Assistant installation, and entity restrictions are opt-in rather than a default limit on device control.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150554,
      "name": "hermes-ghost-call",
      "does": "Runs a live call copilot that transcribes conversation, offers private hints and earpiece speech, and produces a post-call summary.",
      "useful": 4,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 4,
      "total": 26,
      "best": "A working audio bridge, private whisper routing, bounded sessions and layered intervention gates go far beyond a conversational skill.",
      "worst": "The README explicitly leaves live Twilio calls and use inside a running Hermes chat untested, and setup needs a separately reachable bridge.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151995,
      "name": "courierpr",
      "does": "Connects Hermes to a public remote MCP server for browsing and searching CourierPR company press releases.",
      "useful": 2,
      "reliable": 2,
      "polish": 3,
      "creative": 1,
      "beyond": 2,
      "total": 12,
      "best": "The small read-only package needs no credentials and clearly labels releases as company statements.",
      "worst": "The pinned repository contains only configuration and a skill, with no server implementation or tests and no completed Hermes testing reported in the PR.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150503,
      "name": "hermes-windows-certificates",
      "does": "Adds two read-only Windows tools that enumerate certificate stores and explain individual certificates, validity and key purposes.",
      "useful": 3,
      "reliable": 4,
      "polish": 3,
      "creative": 2,
      "beyond": 3,
      "total": 18,
      "best": "Direct crypt32 integration, explicit pointer signatures and unit/live test cases give useful diagnostics without shell parsing or elevation.",
      "worst": "The self_signed field only compares subject and issuer names, so it does not verify that a certificate actually signed itself.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 149534,
      "name": "cronforge",
      "does": "Observes recurring Hermes cron executions and generates model-free shell replay scripts with activation and rollback tools.",
      "useful": 4,
      "reliable": 2,
      "polish": 4,
      "creative": 4,
      "beyond": 4,
      "total": 22,
      "best": "Reducing repeated model work through recorded execution evidence and scheduler integration directly improves the user’s agent.",
      "worst": "Command flattening discards working directories and timeouts, and generated scripts assume exit zero rather than preserving recorded exit assertions, weakening the claimed equivalence.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 149147,
      "name": "camel-guard",
      "does": "Gates sensitive Hermes tool calls after untrusted output using capability categories classified solely from the current user request.",
      "useful": 5,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 27,
      "best": "Native hooks enforce a code-level authority boundary with fail-closed classification and tests covering direct, concurrent and sequential dispatch.",
      "worst": "Authorization is only by broad effect category, so an already-authorized capability can still be misused for a different path, recipient or command.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 151681,
      "name": "overhead",
      "does": "Measures vegetation, burn and water changes from Sentinel-2 imagery and delivers comparison images, timelapses and interactive Desktop cards.",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 5,
      "beyond": 4,
      "total": 25,
      "best": "A substantive geospatial engine computes pixel-based measurements with provenance, cloud handling and tests, then presents them through polished Hermes media surfaces.",
      "worst": "Spectral thresholds can confuse seasonal change or land clearing with damage, and the image cache grows without automatic pruning.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 152772,
      "name": "hermes-superagent",
      "does": "Adds a gated coordinator, isolated task workers, a progress watchdog and a persistent ledger fed by external task and session connectors.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "best": "Role tool guards, stale-worker fencing, bounded repairs and completeness-aware ledger tests materially improve multi-agent supervision.",
      "worst": "Real-model coverage is narrow: watchdog repairs have not run on real Hermes, and Gmail, Calendar and WhatsApp connectors have only synthetic-data coverage.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152585,
      "name": "othigai",
      "does": "Runs consented prompt-injection drills against Hermes subagents using fake secrets, controlled tools and a local listener to measure attempted leaks.",
      "useful": 4,
      "reliable": 3,
      "polish": 4,
      "creative": 5,
      "beyond": 5,
      "total": 26,
      "best": "Paired clean controls, real-Hermes stub-model tests and lifecycle containment make agent safety measurable on the user’s own setup.",
      "worst": "Known grading gaps can misclassify harmless payloads as failures or missed sends as resistance, undermining the central safety score.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151272,
      "name": "hermes-helm",
      "does": "Adds owner-only Discord and Telegram controls for Hermes tools, skills, MCP servers, lost-turn replay, long replies and daily spending limits.",
      "useful": 5,
      "reliable": 4,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "total": 26,
      "best": "It brings practical control of a continuously running agent to the user’s phone, with tested owner gates and explicit replay permissions.",
      "worst": "The spending gate counts only priced calls, so unknown model prices escape the limit unless the owner supplies rates.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 151501,
      "name": "hermes-pentest",
      "does": "Prepares and runs bounded local application security assessments with human approval, isolated runtimes, durable evidence and regression checks.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "best": "Exact-scope one-use approval, default-deny unmanaged targets and controlled SQL/XSS tests make it a substantial extension of Hermes execution.",
      "worst": "The guided workflow supports a limited Vite/Supabase recipe and needs prepared local Docker images and browser tooling rather than offering general application coverage.",
      "standout": false,
      "blocked": null
    }
  ]
}