{
  "batch": "read1-044",
  "plugins": [
    {
      "pr": 0,
      "name": "done-bell",
      "does": "Notifies users when Hermes finishes a turn or awaits approval, with a Desktop status chip and mute controls.",
      "useful": 4,
      "reliable": 4,
      "polish": 3,
      "creative": 2,
      "beyond": 4,
      "best": "Native turn hooks and Desktop busy-state watching make a small, immediately practical improvement to the user's own agent.",
      "worst": "Installation instructions use the author's Windows development path and require manually copying the Desktop pane; tests cover Python state and hooks but not actual Desktop notifications.",
      "total": 21,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "does": "Connects Hermes to Home Assistant through filtered device events, REST control tools and notification delivery.",
      "useful": 4,
      "reliable": 5,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "best": "A genuine gateway platform combines event subscriptions, reconnect backoff, cron delivery and tested entity and service restrictions.",
      "worst": "Its value requires an existing Home Assistant installation and a powerful access token, with entity restrictions opt-in.",
      "total": 26,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153919,
      "name": "hermes-workflow",
      "does": "Adds a Desktop graph editor and tools for reusable Python nodes and saved executable workflows.",
      "useful": 4,
      "reliable": 2,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "best": "A substantial native visual editor, reusable node library and importable illustrated demos make workflows inspectable inside Hermes Desktop.",
      "worst": "The add_node_to_workflow handler incorrectly applies the config validator to execution data, and the backend executes arbitrary node Python without isolation or a timeout; no tests are shipped.",
      "total": 23,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153103,
      "name": "bsc-mint-guard",
      "does": "Reads BSC contract metadata, mint conditions and wallet balances to produce free-mint preview checks and price quotes.",
      "useful": 2,
      "reliable": 3,
      "polish": 3,
      "creative": 2,
      "beyond": 3,
      "best": "Read-only standard-library tools avoid private keys and spending, with mocked tests for verdicts, quantities and RPC fallback.",
      "worst": "Fixed selector guesses cannot establish mint safety, failed wallet reads still return already_minted=false, and sequential RPC fallback has no overall deadline.",
      "total": 16,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151433,
      "name": "linkzoid",
      "does": "Bundles instructions for using a separately configured hosted LinkZoid MCP service to audit and approve website internal links.",
      "useful": 3,
      "reliable": 2,
      "polish": 3,
      "creative": 2,
      "beyond": 1,
      "best": "Clear setup and troubleshooting documentation explains OAuth, analysis-credit costs and confirmation before live-site changes.",
      "worst": "The pinned entry is just a manifest and skill, with no connector implementation or tests; its claimed functionality and approval enforcement depend on an external service and model instructions.",
      "total": 12,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152559,
      "name": "psychopomp",
      "does": "Keeps approved weekly journal pages and sealed letters, then delivers them after prolonged silence and a witness code.",
      "useful": 3,
      "reliable": 3,
      "polish": 5,
      "creative": 5,
      "beyond": 5,
      "best": "The native journal cards, witness workflow and rehearsal using the real state machine on a virtual clock form an unusually coherent agent-native product.",
      "worst": "Check-in state advances before send success and Book delivery has no persisted retry state, weakening its legacy-delivery promise; letters are also stored in plaintext.",
      "standout": true,
      "total": 26,
      "blocked": null
    },
    {
      "pr": 149920,
      "name": "tightbeam",
      "does": "Audits and optimizes YouTube catalogs and uploads, schedules and monitors multi-platform shorts through OpusClip.",
      "useful": 3,
      "reliable": 3,
      "polish": 4,
      "creative": 3,
      "beyond": 4,
      "best": "Host-owned OAuth and MCP credentials, resumable catalog sweeps and automation blueprints combine into a useful creator workflow.",
      "worst": "Scheduling treats a failed collision lookup as an empty calendar and counts writes as scheduled without matching read-back entries; no regression tests are shipped in the pinned tree.",
      "total": 21,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153934,
      "name": "adeptory",
      "does": "Builds dedicated Hermes specialist profiles with searchable source libraries, procedures, tool capability records and retained verification evidence.",
      "useful": 4,
      "reliable": 4,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "best": "Profile delivery is gated on revision-bound native task evidence, with source-passage checks, library freshness tracking and substantial mechanics and integration tests.",
      "worst": "Specialist quality still depends on extensive model-directed research and host tool execution, while passage matching and receipts cannot establish the semantic correctness of professional advice.",
      "standout": false,
      "total": 26,
      "blocked": null
    },
    {
      "pr": 153760,
      "name": "a11y-lens",
      "does": "Runs axe-core in real Chromium, stores accessibility evidence and compares scans to verify repairs and identify regressions.",
      "useful": 4,
      "reliable": 5,
      "polish": 5,
      "creative": 3,
      "beyond": 4,
      "best": "A complete scan-repair-rescan loop pairs isolated browser execution and DNS pinning with evidence-aware comparison, self-tests and readable reports.",
      "worst": "It audits one unauthenticated page at a time, and blocking other hosts can leave remote pages partially rendered; a clean automated scan does not establish accessibility conformance.",
      "standout": false,
      "total": 25,
      "blocked": null
    },
    {
      "pr": 151674,
      "name": "numogram",
      "does": "Connects Hermes to ESP32 boards to watch crashes, decode ELF source locations, check pins, build firmware and flash with approval.",
      "useful": 3,
      "reliable": 5,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "best": "A real hardware feedback loop combines persistent serial watching, crash-triggered agent wakeups, deterministic decoding and code-enforced build/flash approval, with a no-board emulator demo and safety tests.",
      "worst": "It supports one board at a time and ESP-IDF projects, with real-hardware coverage concentrated on ESP32-P4 and virtual ESP32-C3 rather than the full advertised chip family.",
      "standout": true,
      "total": 27,
      "blocked": null
    },
    {
      "pr": 152203,
      "name": "whatsapp-plus",
      "does": "Extends Hermes WhatsApp with persistent group context, trust tiers, messaging tools, campaigns and optional voice calls.",
      "useful": 4,
      "reliable": 1,
      "polish": 3,
      "creative": 3,
      "beyond": 5,
      "best": "Persistent group context and trust-tier hooks address practical limits of using a personal agent in busy chats.",
      "worst": "The pinned code re-registers the bundled WhatsApp platform, rebuilds its bridge at runtime and directly reads .env; these conflict with catalog requirements, and the maintainer explicitly declined the listing.",
      "total": 21,
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150778,
      "name": "werracle-evm-guard",
      "does": "Intercepts selected cast send terminal commands with EVM approval/value risk heuristics and exposes hosted Radar telemetry tools.",
      "useful": 3,
      "reliable": 2,
      "polish": 3,
      "creative": 2,
      "beyond": 4,
      "best": "A real pre-tool blocking hook handles raw calldata, readable signatures and nested shell commands, while the API key uses Hermes secret scoping.",
      "worst": "This is heuristic parsing rather than EVM execution simulation: the hook hardcodes Base, misses other broadcast routes and unresolved shell variables, and uses a chain-independent trusted-address list.",
      "total": 18,
      "standout": false,
      "blocked": null
    }
  ]
}