{
  "batch": "read1-009",
  "plugins": [
    {
      "pr": 0,
      "name": "done-bell",
      "does": "Notifies the user when Hermes finishes a turn or needs approval, with a chime, OS notifications, and a Desktop status pane.",
      "useful": 4,
      "reliable": 3,
      "polish": 3,
      "creative": 2,
      "beyond": 4,
      "total": 20,
      "best": "Real lifecycle hooks and Desktop busy-state observation provide useful completion and approval notifications without credentials or an external service.",
      "worst": "Setup still uses the author's local Windows installation path and a manual Desktop-file copy, while the shipped tests cover backend state rather than actual Desktop notifications.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "does": "Connects Hermes to Home Assistant through four device tools, filtered state-change events, and notification delivery including standalone cron jobs.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "total": 25,
      "best": "The bidirectional gateway and device tools use scoped credentials, reconnect handling, explicit entity policies, and substantial regression tests.",
      "worst": "Entity restrictions are opt-in, leaving broad device-control authority until the operator configures allowlists or denylists.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152450,
      "name": "heats",
      "does": "Runs checks before Hermes finishes coding and can race subagent fixes in git worktrees, rank their tested patches, and apply a selected result.",
      "useful": 5,
      "reliable": 4,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "total": 28,
      "best": "The pinned implementation combines test-driven continuation, guarded worktree races, human approval for untrusted checks, and a useful Desktop scoreboard with extensive regression coverage.",
      "worst": "Isolation is incomplete: dependency directories are shared with the real checkout and dynamically computed shell paths can escape the worktree, while shell-only edits do not trigger the referee.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 150429,
      "name": "plan-tracker",
      "does": "Turns markdown checklists into a Desktop progress pane with milestones, history, effort estimates, and editable task checkboxes.",
      "useful": 4,
      "reliable": 1,
      "polish": 4,
      "creative": 3,
      "beyond": 4,
      "total": 20,
      "best": "It makes existing markdown plans usable as a session-aware progress interface, with content-hash checks on checkbox edits and extensive parser and Desktop tests.",
      "worst": "At the pinned commit, POST and DELETE /block pass unchecked IDs into JSON filenames, allowing path traversal and modification of files outside the plugin data directory.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152576,
      "name": "openchi",
      "does": "Offers illustrated stretching and breathing sessions in Hermes Desktop during long tasks, with local message-tension detection and user-controlled timers.",
      "useful": 3,
      "reliable": 5,
      "polish": 5,
      "creative": 4,
      "beyond": 4,
      "total": 25,
      "best": "Its carefully tested Desktop experience waits for a user click, handles cooldowns and sleep recovery, and keeps message text out of storage and network requests.",
      "worst": "Its usefulness is concentrated in Desktop wellness breaks, with no additional capability for executing the agent's tasks.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153844,
      "name": "fine-print",
      "does": "Observes Hermes tool results and model requests to expose information loss through targeted model notes, end-of-turn notices, and local limit reports.",
      "useful": 5,
      "reliable": 4,
      "polish": 4,
      "creative": 5,
      "beyond": 5,
      "total": 28,
      "best": "It gives users and models concrete evidence of otherwise hidden truncation and image loss through native hooks, bounded annotations, and extensive regression fixtures.",
      "worst": "Detectors depend on Hermes 0.23.0 formats, and the README reports that Desktop notices fail on that release, leaving notes and slash-command reports as fallbacks.",
      "standout": true,
      "blocked": null
    },
    {
      "pr": 149478,
      "name": "contract-hunter",
      "does": "Wraps Scout commands as Hermes tools and slash commands for finding EVM bounty targets, fetching source, running fork proofs, and producing reports.",
      "useful": 3,
      "reliable": 2,
      "polish": 3,
      "creative": 2,
      "beyond": 3,
      "total": 16,
      "best": "The wrapper uses argument lists, subprocess timeouts, a minimal child environment, and a default-off session-start pull.",
      "worst": "The documented proof workflow passes --hypothesis to scout prove, but the Scout revision named by the plugin accepts only an ID there and exposes hypothesis as a separate subcommand; no plugin tests catch this mismatch.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150349,
      "name": "lampstand",
      "does": "Maintains shared project memory and rules in a git repository, with native task tools, subagent context, a Desktop cockpit, and optional rooms and scheduled jobs.",
      "useful": 4,
      "reliable": 1,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 23,
      "best": "Shared task records, git history, constitution projections, and room wakeups connect persistent user context across sessions, subagents, and Desktop.",
      "worst": "The pinned release reads Hermes .env directly, forwards the gateway environment to external Agora code, and lets newly edited scheduled jobs run without renewed operator approval.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150954,
      "name": "meme-studio-mcp",
      "does": "Registers the hosted Meme Studio MCP service and a skill for selecting templates, captioning memes, and returning images or editor links.",
      "useful": 3,
      "reliable": 3,
      "polish": 4,
      "creative": 2,
      "beyond": 2,
      "total": 16,
      "best": "Clear installation, an illustrated demo, explicit remote-service disclosures, and an untrusted MCP declaration make the narrow connector easy to understand.",
      "worst": "The pinned submission contains configuration and a skill only, so rendering and runtime tool behavior depend entirely on a closed-source hosted service outside the reviewed commit.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152906,
      "name": "sochen",
      "does": "Connects Hermes to Sochen for cloud jobs and US business phone calls, including mid-call answers, job updates, and a Desktop desk.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 4,
      "total": 24,
      "best": "Native credit-spending approval, scoped secrets, bounded job tracking, and tested Desktop and phone-call flows make this more substantial than a simple API wrapper.",
      "worst": "The response shaper reports a reply with no status as ok:true, so a malformed service response can appear to be an accepted job.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151277,
      "name": "developh",
      "does": "Connects a hosted website-audit MCP service and a skill that turns PageSpeed findings into ranked fixes, with optional studio inquiries.",
      "useful": 3,
      "reliable": 3,
      "polish": 3,
      "creative": 1,
      "beyond": 1,
      "total": 12,
      "best": "The README clearly explains installation, the audit workflow, data destinations, and consent for inquiries, and the manifest marks the remote server untrusted.",
      "worst": "The package contains no audit implementation or tests, and its own skill already calls the same service directly over HTTP without the plugin.",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151576,
      "name": "hermes-witness",
      "does": "Grades newly written Hermes skills against tool outcomes and user constraints, quarantines weak skills, and records later usage outcomes in a hash-chained ledger.",
      "useful": 4,
      "reliable": 2,
      "polish": 3,
      "creative": 4,
      "beyond": 5,
      "total": 23,
      "best": "Native hooks connect skill creation and later use to a local evidence trail, with reversible quarantine and regression tests for the main grading and recovery paths.",
      "worst": "The error matcher treats \"0 failed\" as failure and the constraint matcher can flag \"never git push main\", so the default gate can quarantine valid skills.",
      "standout": false,
      "blocked": null
    }
  ]
}