{
  "batch": "read1-137",
  "plugins": [
    {
      "pr": 0,
      "name": "done-bell",
      "does": "Adds completion and approval notifications, an optional chime, and a Desktop status chip. (Source: https://raw.githubusercontent.com/aydnOktay/hermes-done-bell-/3fa519a291d91b0e6c44bd42762190363ce80c39/README.md)",
      "useful": 4,
      "reliable": 3,
      "polish": 2,
      "creative": 2,
      "beyond": 4,
      "total": 19,
      "best": "Lifecycle hooks and a Desktop busy-state watcher cover both terminal and gateway activity without retaining approval commands. (Source: https://raw.githubusercontent.com/aydnOktay/hermes-done-bell-/3fa519a291d91b0e6c44bd42762190363ce80c39/desktop/plugin.js; https://raw.githubusercontent.com/aydnOktay/hermes-done-bell-/3fa519a291d91b0e6c44bd42762190363ce80c39/README.md)",
      "worst": "The README gives a personal Windows development path and manual pane-copy setup, while the tests cover only backend behavior. (Source: https://raw.githubusercontent.com/aydnOktay/hermes-done-bell-/3fa519a291d91b0e6c44bd42762190363ce80c39/desktop/plugin.js; https://raw.githubusercontent.com/aydnOktay/hermes-done-bell-/3fa519a291d91b0e6c44bd42762190363ce80c39/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 0,
      "name": "homeassistant",
      "does": "Connects Hermes to Home Assistant state-change events, device-query and control tools, and notification delivery. (Source: https://raw.githubusercontent.com/NousResearch/hermes-official-plugins/e7a218f26db88426ab3e7376f0e000b6e356f5e7/homeassistant/README.md)",
      "useful": 4,
      "reliable": 5,
      "polish": 4,
      "creative": 3,
      "beyond": 5,
      "total": 26,
      "best": "A genuine gateway platform combines filtered event subscriptions with scoped credentials, service restrictions, reconnect handling, and extensive policy tests. (Source: https://raw.githubusercontent.com/NousResearch/hermes-official-plugins/e7a218f26db88426ab3e7376f0e000b6e356f5e7/homeassistant/tools.py; https://raw.githubusercontent.com/NousResearch/hermes-official-plugins/e7a218f26db88426ab3e7376f0e000b6e356f5e7/homeassistant/README.md)",
      "worst": "Entity restrictions are opt-in, and scripts or scenes can have effects beyond their named entity; operators must carefully scope their Home Assistant access. (Source: https://raw.githubusercontent.com/NousResearch/hermes-official-plugins/e7a218f26db88426ab3e7376f0e000b6e356f5e7/homeassistant/tools.py)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 154011,
      "name": "disko",
      "does": "Controls a separately installed disko AI DJ app through tools, slash commands, a CLI, and live now-playing context. (Source: https://raw.githubusercontent.com/emirkaan5/hermes-disko/92941008c78981e935af6744e09ce2d441056e48/README.md)",
      "useful": 3,
      "reliable": 4,
      "polish": 4,
      "creative": 3,
      "beyond": 3,
      "total": 20,
      "best": "Real Unix-socket integration tests cover playback controls, safe argument construction, launch failures, and refusing to quit a process the plugin did not start. (Source: https://raw.githubusercontent.com/emirkaan5/hermes-disko/92941008c78981e935af6744e09ce2d441056e48/README.md)",
      "worst": "The substantial DJ engine and Mac setup are external prerequisites, and optional remote control sends its bearer token over plain HTTP. (Source: https://raw.githubusercontent.com/emirkaan5/hermes-disko/92941008c78981e935af6744e09ce2d441056e48/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153330,
      "name": "waifu",
      "does": "Displays a Live2D dashboard companion whose visual activity follows Hermes lifecycle hooks and a user-selected live session. (Source: https://raw.githubusercontent.com/waifuai/hermes-waifu/173bd90f15f0de9ddf5cb172f6df5939b34b5ad5/README.md)",
      "useful": 2,
      "reliable": 2,
      "polish": 4,
      "creative": 2,
      "beyond": 4,
      "total": 18,
      "best": "Bounded hashed session state and an isolated renderer keep conversation payloads out of the external asset runtime. (Source: https://raw.githubusercontent.com/waifuai/hermes-waifu/173bd90f15f0de9ddf5cb172f6df5939b34b5ad5/README.md)",
      "worst": "The README explicitly says rights for 52 character models and thumbnails are unestablished and remain a catalog-review blocker; rendering also depends on CDN availability. (Source: https://raw.githubusercontent.com/waifuai/hermes-waifu/173bd90f15f0de9ddf5cb172f6df5939b34b5ad5/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 151939,
      "name": "bluff",
      "does": "Benchmarks a Hermes model on 24 real and fabricated functions, exceptions, and elements, saving graded reports and model comparisons. (Source: https://raw.githubusercontent.com/salch-cred/bluff/4c43c96be7a4411d91ebc0b7de59086639ce0507/README.md)",
      "useful": 3,
      "reliable": 4,
      "polish": 3,
      "creative": 3,
      "beyond": 3,
      "total": 19,
      "best": "Balanced real-item controls, deterministic sampling, ground-truth tests, and fail-fast provider handling make the narrow benchmark auditable. (Source: https://raw.githubusercontent.com/salch-cred/bluff/4c43c96be7a4411d91ebc0b7de59086639ce0507/judge.py; https://raw.githubusercontent.com/salch-cred/bluff/4c43c96be7a4411d91ebc0b7de59086639ce0507/README.md)",
      "worst": "Scoring trusts the explicit exists field rather than checking whether the answer text fabricates details, so this is a narrow lower-bound probe rather than a general hallucination measure. (Source: https://raw.githubusercontent.com/salch-cred/bluff/4c43c96be7a4411d91ebc0b7de59086639ce0507/judge.py)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150323,
      "name": "coldstart",
      "does": "Retrieves tiered document evidence and uses Hermes input and held-output hooks to flag or withhold unsupported quotations, figures, and identifiers. (Source: https://raw.githubusercontent.com/zvporter-ops/coldstart/2f8d036e070f7a4bc593ab9fa5b60dcdd820cab3/README.md)",
      "useful": 4,
      "reliable": 3,
      "polish": 4,
      "creative": 4,
      "beyond": 5,
      "total": 25,
      "best": "The output-transform hook can prevent unchecked text from streaming, and the implementation distinguishes internal from public evidence with locality checks on model access. (Source: https://raw.githubusercontent.com/zvporter-ops/coldstart/2f8d036e070f7a4bc593ab9fa5b60dcdd820cab3/__init__.py; https://raw.githubusercontent.com/zvporter-ops/coldstart/2f8d036e070f7a4bc593ab9fa5b60dcdd820cab3/README.md)",
      "worst": "Strict mode still allows unsupported prose when it lacks a checked quote, figure, date, URL, or identifier, so the opening promise that every answer is grounded or refused is broader than the mechanical gate. (Source: https://raw.githubusercontent.com/zvporter-ops/coldstart/2f8d036e070f7a4bc593ab9fa5b60dcdd820cab3/__init__.py; https://raw.githubusercontent.com/zvporter-ops/coldstart/2f8d036e070f7a4bc593ab9fa5b60dcdd820cab3/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152775,
      "name": "launch-desk",
      "does": "Builds an approved launch pack with ranked post references, mechanically checked copy, branded images, optional video, and consented Hermes check-ins. (Source: https://raw.githubusercontent.com/posts-design/hermes-launch-desk/745208481188ebad236eb5dc8f9291fecc7aad22/README.md)",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 3,
      "beyond": 4,
      "total": 23,
      "best": "The implemented workflow preserves the founder's own words, exposes failed checks, exports usable media, and schedules reminders only after an explicit Yes. (Source: https://raw.githubusercontent.com/posts-design/hermes-launch-desk/745208481188ebad236eb5dc8f9291fecc7aad22/launch_desk/wizard.py; https://raw.githubusercontent.com/posts-design/hermes-launch-desk/745208481188ebad236eb5dc8f9291fecc7aad22/README.md)",
      "worst": "Its niche value depends on posts.design for fresh references and a separate renderer for finished video; deterministic copy checks cannot establish that every prose claim is true. (Source: https://raw.githubusercontent.com/posts-design/hermes-launch-desk/745208481188ebad236eb5dc8f9291fecc7aad22/launch_desk/wizard.py; https://raw.githubusercontent.com/posts-design/hermes-launch-desk/745208481188ebad236eb5dc8f9291fecc7aad22/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 153370,
      "name": "game-lab",
      "does": "Bundles safe-copy game modding, RPG Maker localization, sprite-sheet checks, and experimental X11 playtesting tools. (Source: https://raw.githubusercontent.com/flabanabba8/game-lab/3173e1b09860be1c4aac3c952cf97b9447cdc55e/README.md)",
      "useful": 3,
      "reliable": 2,
      "polish": 3,
      "creative": 3,
      "beyond": 3,
      "total": 17,
      "best": "Revision copies, rollback, localization control-code checks, and approval-gated input are concrete safeguards for users editing their games. (Source: https://raw.githubusercontent.com/flabanabba8/game-lab/3173e1b09860be1c4aac3c952cf97b9447cdc55e/README.md)",
      "worst": "The pinned README reports 0/4 complete correct SuperTux batch runs and 0/2 Packet Storm runs, while the parts share no data and most end-to-end game workflows remain unverified. (Source: https://raw.githubusercontent.com/flabanabba8/game-lab/3173e1b09860be1c4aac3c952cf97b9447cdc55e/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 150900,
      "name": "ferry",
      "does": "The PR describes asynchronous job exchange between Hermes instances over a shared Git repository, but its implementation was unreadable. (PR description: https://api.github.com/repos/NousResearch/hermes-agent/pulls/150900)",
      "useful": 0,
      "reliable": 0,
      "polish": 0,
      "creative": 0,
      "beyond": 0,
      "total": 0,
      "best": "The PR proposes an audit-friendly store-and-forward transport for intermittently connected agents; this could not be verified in code. (PR description: https://api.github.com/repos/NousResearch/hermes-agent/pulls/150900)",
      "worst": "The pinned source could not be inspected, so the PR's capability and test claims are unverified. (Pinned repository: https://github.com/teal-sea/ferry; pin record: https://gist.githubusercontent.com/gtmcknight/594c47336546c8763e641b46cf9af4ec/raw/field.json)",
      "standout": false,
      "blocked": "The pinned repository https://github.com/teal-sea/ferry and pinned raw README return HTTP 404, and git fetch fails requesting GitHub authentication."
    },
    {
      "pr": 152547,
      "name": "gambit",
      "does": "Adds chess play and coaching through a terminal board, chat tools, and Desktop, with local engine analysis and public-game replay. (Source: https://raw.githubusercontent.com/itsflownium/gambit-hackathon/9c725ad0b04c2bc32f16efd00dfc1b4d421360cb/README.md)",
      "useful": 3,
      "reliable": 4,
      "polish": 5,
      "creative": 3,
      "beyond": 4,
      "total": 23,
      "best": "The code validates legal moves, guards against stale asynchronous replies, and tests a deadline fallback so a stalled model cannot make a second late move. (Source: https://raw.githubusercontent.com/itsflownium/gambit-hackathon/9c725ad0b04c2bc32f16efd00dfc1b4d421360cb/README.md)",
      "worst": "It needs a separately installed Stockfish and a working model for conversational coaching, and the documented verified catalog scope is macOS only. (Source: https://raw.githubusercontent.com/itsflownium/gambit-hackathon/9c725ad0b04c2bc32f16efd00dfc1b4d421360cb/README.md)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 149438,
      "name": "nurse",
      "does": "Lists pm2 and systemd services, diagnoses log fingerprints, and performs approval-gated service actions with persisted cooldowns. (Source: https://raw.githubusercontent.com/Megumiiiiii/hermes-nurse/a830f7c8a71490c69368cf2071b85b4476e561b0/README.md)",
      "useful": 4,
      "reliable": 3,
      "polish": 3,
      "creative": 2,
      "beyond": 3,
      "total": 18,
      "best": "State-changing repairs request Hermes human approval in addition to a plan/confirm step, allowlisted verbs, and a persistent repair budget. (Source: https://raw.githubusercontent.com/Megumiiiiii/hermes-nurse/a830f7c8a71490c69368cf2071b85b4476e561b0/hermes_nurse/repair.py; https://raw.githubusercontent.com/Megumiiiiii/hermes-nurse/a830f7c8a71490c69368cf2071b85b4476e561b0/README.md)",
      "worst": "The crash-loop rate divides the cumulative pm2 restart count by the current process uptime, so historical restarts can still falsely block a healthy service; ledger updates also lack locking. (Source: https://raw.githubusercontent.com/Megumiiiiii/hermes-nurse/a830f7c8a71490c69368cf2071b85b4476e561b0/hermes_nurse/repair.py)",
      "standout": false,
      "blocked": null
    },
    {
      "pr": 152948,
      "name": "local-fleet",
      "does": "Routes Hermes primary inference, auxiliary tasks, and subagents across registered local servers with cloud fallback, private mode, and a Desktop fleet console. (Source: https://raw.githubusercontent.com/humedini/local-fleet/94cd17d2477f9d1a6f39aeee153be1d25c5b57e2/README.md)",
      "useful": 4,
      "reliable": 3,
      "polish": 5,
      "creative": 4,
      "beyond": 5,
      "total": 26,
      "best": "It meaningfully changes where Hermes runs inference, with shared eligibility rules, reversible configuration, drift reporting, privacy controls, and substantial behavioral tests. (Source: https://raw.githubusercontent.com/humedini/local-fleet/94cd17d2477f9d1a6f39aeee153be1d25c5b57e2/health.py; https://raw.githubusercontent.com/humedini/local-fleet/94cd17d2477f9d1a6f39aeee153be1d25c5b57e2/README.md)",
      "worst": "Health probes submit workers to ThreadPoolExecutor without copying context, yet resolve scoped secrets inside those workers, risking failed or incorrectly scoped authentication in multiplexed profiles. (Source: https://raw.githubusercontent.com/humedini/local-fleet/94cd17d2477f9d1a6f39aeee153be1d25c5b57e2/health.py)",
      "standout": true,
      "blocked": null
    }
  ]
}
