{ "batch": "read1-071", "plugins": [ { "pr": 0, "name": "done-bell", "does": "Rings an OS notification, optional chime and a Desktop status chip when a Hermes turn finishes or an approval is waiting.", "useful": 3, "reliable": 3, "polish": 3, "creative": 2, "beyond": 3, "total": 17, "best": "Small, careful hook set (on_session_end, approval hooks) with never-raise handlers, a Desktop pane, a /bell command and tests.", "worst": "A modest convenience feature with only a handful of unit tests, and its README install steps are hard-coded to the author's Windows paths.", "standout": false, "blocked": null }, { "pr": 0, "name": "homeassistant", "does": "Official Home Assistant gateway platform plus four tools to list, read and control smart-home entities over HA's REST and WebSocket APIs.", "useful": 4, "reliable": 5, "polish": 4, "creative": 3, "beyond": 5, "total": 26, "best": "A full gateway platform adapter with entity allow/deny lists, a service policy, password-flagged secrets and about 2,300 lines of tests.", "worst": "No Desktop UI or screenshots, and the idea (an HA bridge) is a well-known integration rather than something new.", "standout": false, "blocked": null }, { "pr": 149529, "name": "upstash", "does": "Memory provider that recalls Upstash Redis Search memories into each turn, stores searchable chat history, mirrors the built-in memory tool and adds four memory tools.", "useful": 4, "reliable": 4, "polish": 4, "creative": 3, "beyond": 5, "total": 25, "best": "A real MemoryProvider that changes Hermes's memory, reads its token through Hermes's get_secret, keeps writes to the primary agent only and has tests for the provider and Redis layout.", "worst": "Needs a hosted Upstash database and the upstash-redis package, which a plain directory install of the catalog entry does not bring in by itself.", "standout": false, "blocked": null }, { "pr": 152749, "name": "teach-back", "does": "Tracks every file the agent writes or patches through post_tool_call and runs /teachback quizzes, written and graded by the user's own model, on what that code does.", "useful": 3, "reliable": 3, "polish": 4, "creative": 4, "beyond": 3, "total": 20, "best": "A fresh idea, paying down 'comprehension debt' by quizzing the user on agent-written code built from the real diffs, with a lines-understood score.", "worst": "The pull request says it is 'Not yet verified with a live model', and the questions and grading rest entirely on LLM prompts.", "standout": false, "blocked": null }, { "pr": 151098, "name": "coffee-vram", "does": "An llm_execution middleware that checks whether a local Ollama model fits in VRAM before each call, queues or evicts idle models, and shows a Desktop chip, pane and /vram.", "useful": 4, "reliable": 4, "polish": 5, "creative": 4, "beyond": 5, "total": 27, "best": "Changes how Hermes runs on the user's own GPU with a real fit estimator, a bounded queue (it times out and runs the call anyway) and many unit and e2e tests, plus measured research.", "worst": "Only helps users of an external Ollama, and its many tuning settings (pool, KV type, overhead) take hardware knowledge to set correctly.", "standout": true, "blocked": null }, { "pr": 153961, "name": "proving-ground", "does": "A /provingground CLI that sends 24 harmless synthetic tool calls through the user's installed plugins in a throwaway child Hermes home, without calling the model, and reports what each plugin and the whole stack blocked, rewrote or let through.", "useful": 3, "reliable": 5, "polish": 5, "creative": 5, "beyond": 4, "total": 26, "best": "A genuinely new agent tool: it finds cross-plugin interactions (a rewriter slipping past a guard) through Hermes's real hook chain, isolated from the user's home, with dozens of test files and a hosted sample report.", "worst": "Its audience is plugin authors and power users with large stacks, it can download about 650 MB of runtime on first use, and the pull request was still open when read.", "standout": true, "blocked": null }, { "pr": 151784, "name": "audionotes", "does": "An Agent Plugins package that connects Hermes through mcp-remote to Audionotes' hosted read-only MCP server (search, list and read voice notes and transcripts) with OAuth and one retrieval skill.", "useful": 2, "reliable": 3, "polish": 3, "creative": 2, "beyond": 3, "total": 16, "best": "Read-only, OAuth through a pinned mcp-remote version, marked 'untrusted', with a clear README on what data leaves and where.", "worst": "No code of its own: a config pointing at a vendor's hosted MCP server that needs a paid Audionotes account, so its value rests entirely on that service.", "standout": false, "blocked": null }, { "pr": 151215, "name": "tuneviva", "does": "An Agent Plugins package that points Hermes at TuneViva's keyless read-only MCP server for sheet-music search and practice-score links, plus two guidance skills.", "useful": 2, "reliable": 3, "polish": 2, "creative": 2, "beyond": 2, "total": 13, "best": "No account or key needed and the endpoint is read-only.", "worst": "No code at all: a remote MCP URL and two skills, with a README of a few lines, a narrow audience, and no tests.", "standout": false, "blocked": null }, { "pr": 150234, "name": "va-claims", "does": "Seven tools and a Desktop page for US veterans: a 38 CFR filing guide, intent-to-file and benefit deadline tracking, and live claim status and contention lookup through the VA Lighthouse APIs.", "useful": 3, "reliable": 2, "polish": 4, "creative": 4, "beyond": 3, "total": 19, "best": "A real, underserved use with live VA Lighthouse API calls, local deadline tracking that works with no setup, and a Desktop benefits page with screenshots.", "worst": "It signs the VA client assertion with an RSA private key read from a file path rather than Hermes's secret store, there are no tests, and the pull request was closed unmerged by its author.", "standout": false, "blocked": null }, { "pr": 149901, "name": "penguin", "does": "Four deterministic tools for IANA time-zone conversion, whole-meeting overlap and saved team settings, with a themed four-clock Desktop planner page.", "useful": 3, "reliable": 4, "polish": 5, "creative": 3, "beyond": 3, "total": 21, "best": "Deterministic time-zone and DST handling that models get wrong, with a bundled tzdb, core and integration tests, and a heavily polished Desktop page with real screenshots.", "worst": "Meeting-time conversion is a well-trodden idea, and much of the release effort goes to animation, artwork and music rather than to the tools.", "standout": false, "blocked": null }, { "pr": 153624, "name": "rosegold", "does": "An Agent Plugins package pointing Hermes at rosegold's hosted OAuth MCP server, which matches Amazon offers and creator codes to the user's purchases taken from their Gmail order emails.", "useful": 2, "reliable": 2, "polish": 2, "creative": 2, "beyond": 3, "total": 14, "best": "No code of its own and no stored credential: Hermes keeps the OAuth token, and the README spells out what order data is sent.", "worst": "A thin config for a commercial rewards service: it asks the agent to send the user's Amazon order items and prices to a third party, its tools run without approval under default trust, and the pull request is still open.", "standout": false, "blocked": null }, { "pr": 153871, "name": "barehand", "does": "A computer_use provider plus a barehand_machine tool that lets Hermes operate a machine with no OS or agent on it (firmware menus, boot loaders, installers) through QEMU QMP or VNC screen and keyboard, with screen-text reading and a flight recorder.", "useful": 3, "reliable": 4, "polish": 5, "creative": 5, "beyond": 5, "total": 27, "best": "It gives Hermes a new place to act, bare machines through their framebuffer, by registering a real computer_use provider, sends destructive actions to the human-approval gate, and ships recorded demo runs with judge evidence.", "worst": "It needs QEMU or a VNC console and suits a niche of infrastructure work, its tests sit outside the catalog subdir, and the pull request was still open when read.", "standout": true, "blocked": null } ] }