{ "batch": "read1-061", "plugins": [ { "pr": 0, "name": "done-bell", "does": "Notifies users when Hermes finishes a turn or needs approval, with a Desktop status chip, chime and settings pane.", "useful": 4, "reliable": 3, "polish": 3, "creative": 2, "beyond": 4, "total": 20, "best": "Small, keyless integration covers both terminal lifecycle hooks and Desktop busy-state changes with mute controls and bounded local event history.", "worst": "Setup requires a manual Windows pane-copy workaround, notification failures are swallowed, and the tests omit the Desktop notification path.", "standout": false, "blocked": null }, { "pr": 0, "name": "homeassistant", "does": "Adds a Home Assistant gateway platform, filtered state-change events, cron notifications and four device-query/control tools.", "useful": 4, "reliable": 5, "polish": 4, "creative": 3, "beyond": 5, "total": 26, "best": "Combines an event-driven platform with scoped secrets, reconnect handling, explicit device filters and substantial service-policy regression tests.", "worst": "Useful mainly to Home Assistant owners, and device control remains broad unless operators configure the optional entity allowlist or denylist.", "standout": false, "blocked": null }, { "pr": 154075, "name": "longevity-coach", "does": "Adds local coaching outcome ledgers, keyword-based stress triage, biometric recovery verification and a weekly coaching automation blueprint.", "useful": 3, "reliable": 3, "polish": 3, "creative": 4, "beyond": 3, "total": 19, "best": "Recovery verification checks next-night biometrics and confounders, with closed-loop fixture tests rather than relying solely on coaching prose.", "worst": "The full wearable-ingestion coaching loop requires separately installed on-disk skills and setup, while the bundled crunch itself mostly combines triage and reports.", "standout": false, "blocked": null }, { "pr": 152920, "name": "coursebook", "does": "Turns Hermes Desktop into a resumable learning workspace with lesson generation, governed checks, evidence records and parent-led course packs.", "useful": 4, "reliable": 4, "polish": 4, "creative": 4, "beyond": 5, "total": 26, "best": "Transactional lesson transitions, nonce/revision fences and backend-only answer keys make learning progress enforceable beyond a tutoring prompt.", "worst": "The README records incomplete live Desktop verification and partially exercised media/pronunciation surfaces, despite a large backend and UI test suite.", "standout": false, "blocked": null }, { "pr": 153518, "name": "werk", "does": "Runs engineering work items through builder/reviewer roles and domain-pack checks, producing a mechanical model package with an explicit gap report.", "useful": 3, "reliable": 3, "polish": 4, "creative": 5, "beyond": 4, "total": 23, "best": "Combines tracked work, adversarial review and validated, hash-checked CAD tool outputs while exposing unmeasured engineering gaps.", "worst": "Its own pinned documentation admits failed reviewer calibration, checks that cannot establish some claimed requirements, and several-million-token demonstration cost.", "standout": false, "blocked": null }, { "pr": 150783, "name": "solverforge-routing", "does": "Plans and repairs multi-stop driving itineraries using a local Rust road-routing solver, a Desktop map and navigation handoff links.", "useful": 4, "reliable": 5, "polish": 5, "creative": 4, "beyond": 4, "total": 26, "best": "Revision-checked trip edits and independently revalidated route constraints are backed by Python, Rust and Desktop tests, with a polished map and QR handoff.", "worst": "A new area still needs external geocoding/road-data services and a separate engine installation, and the result estimates driving without live traffic.", "standout": true, "blocked": null }, { "pr": 153027, "name": "ceo-pack", "does": "Returns business-role briefs, assignments, decision records, model suggestions and advisory tool-call validation plans.", "useful": 2, "reliable": 3, "polish": 3, "creative": 2, "beyond": 1, "total": 12, "best": "The envelope validator rejects nested credential material and the result verifier distinguishes a proposed call from completed work, with focused offline tests.", "worst": "Despite the orchestration branding, tools return templates and flags without dispatching work, persisting decisions or creating agents, so a skill covers most of the value.", "standout": false, "blocked": null }, { "pr": 149246, "name": "hermes-windows-apps", "does": "Reads Windows Uninstall registry entries to list installed software and report versions, installation details and uninstaller classifications.", "useful": 3, "reliable": 4, "polish": 3, "creative": 2, "beyond": 3, "total": 18, "best": "A compact read-only registry implementation avoids executing uninstallers and includes fake-registry, non-Windows import and live-test coverage.", "worst": "It only inventories registry-registered Windows software, and same-name entries collapse into a first-view summary that can hide version differences.", "standout": false, "blocked": null }, { "pr": 152754, "name": "sluice-search", "does": "Replaces Hermes native web-search and extraction backends with a free-first multi-provider router, persistent health/cache state and a Desktop manager.", "useful": 5, "reliable": 4, "polish": 4, "creative": 3, "beyond": 5, "total": 26, "best": "Improves the existing native web tools across Hermes surfaces with cost-tier fallback, circuit breakers, scoped provider secrets and substantial routing/extraction tests.", "worst": "Reliability still depends on changing public keyless endpoints and optional browser infrastructure, and the README retains stale rename examples and an incorrect API mount string.", "standout": true, "blocked": null }, { "pr": 149819, "name": "umami-digest", "does": "Fetches yesterday's self-hosted Umami analytics and delivers a deterministic morning letter through a no-model Hermes cron job.", "useful": 3, "reliable": 4, "polish": 3, "creative": 2, "beyond": 3, "total": 18, "best": "Handles pagination, unknown-versus-zero traffic, scoped secrets, bounded HTTP reads and schedule recovery with focused tests and no model cost.", "worst": "It serves a narrow analytics niche, requires manual environment setup, and uninstalling before unscheduling can leave a cron job and script behind.", "standout": false, "blocked": null }, { "pr": 153090, "name": "retouch", "does": "Adds a Desktop image-marking workspace where Hermes edits marked regions, keeps per-mark review threads and saves reversible versions.", "useful": 4, "reliable": 5, "polish": 5, "creative": 4, "beyond": 5, "total": 28, "best": "Enforces zero changed pixels outside each mark, rejects invalid version commits and pairs rich Desktop controls with region, hardening and concurrency tests.", "worst": "Generative results still depend on the configured image provider, and hidden editing sessions retain ordinary profile tools while picture-text injection is addressed only by instruction.", "standout": true, "blocked": null }, { "pr": 153186, "name": "language-lock", "does": "Detects reply-language drift locally and injects a correction on the next turn, with per-session language settings and status commands.", "useful": 4, "reliable": 3, "polish": 3, "creative": 3, "beyond": 4, "total": 21, "best": "Adds automatic session-aware monitoring through pre/post-LLM hooks without network calls, dependencies or extra model usage.", "worst": "It cannot repair the current streamed reply, and script/common-word detection plus broad translation-request matching can miss or misclassify multilingual turns.", "standout": false, "blocked": null } ] }