{ "batch": "read1-083", "plugins": [ { "pr": 0, "name": "done-bell", "does": "Notifies when Hermes finishes a turn or needs approval, with a Desktop status chip, chime and mute controls.", "useful": 4, "reliable": 3, "polish": 3, "creative": 2, "beyond": 4, "best": "Desktop busy-state observation and approval hooks provide useful notifications without exposing conversation content.", "worst": "Setup uses the author's local Windows path and a manual Desktop copy, while tests suppress OS notifications and do not cover the Desktop watcher.", "standout": false, "blocked": null, "total": 20 }, { "pr": 0, "name": "homeassistant", "does": "Connects Hermes to Home Assistant events, notifications and four tools for querying and controlling smart-home devices.", "useful": 4, "reliable": 4, "polish": 4, "creative": 3, "beyond": 5, "best": "A real gateway adapter combines filtered events, reconnection, cron delivery and entity-aware control restrictions with extensive regression tests.", "worst": "Entity restrictions are opt-in and cannot constrain what an allowed script or automation does internally, so safe control still depends on the operator's Home Assistant setup.", "standout": false, "blocked": null, "total": 25 }, { "pr": 151732, "name": "agent-latency-budget", "does": "Records recent tool durations per session and exposes budget checks and percentile reports.", "useful": 3, "reliable": 3, "polish": 3, "creative": 2, "beyond": 3, "best": "Monotonic fallback timing and a Hermes-path demo make tool latency observable with little setup and no external services.", "worst": "The manifest promises fail-fast enforcement but checks only completed calls, and per-session dictionaries have no automatic session eviction.", "standout": false, "blocked": null, "total": 17 }, { "pr": 149640, "name": "hermes-ark", "does": "Automatically checkpoints Hermes memory and skills and provides encrypted, versioned full-home backups with selective restore and off-site sync.", "useful": 5, "reliable": 4, "polish": 5, "creative": 4, "beyond": 5, "total": 28, "best": "Staged integrity-checked restores, credential exclusions and extensive adversarial tests support a practical safety net for an agent that modifies its own memory.", "worst": "Restore commits files sequentially after staging, so a failure during final replacement can still leave a partial restore despite the broad atomicity wording.", "standout": true, "blocked": null }, { "pr": 149263, "name": "hermes-phone", "does": "Places user-approved outbound AI phone calls and relays mid-call questions, results and transcripts back to Hermes chat.", "useful": 4, "reliable": 4, "polish": 4, "creative": 4, "beyond": 5, "total": 26, "best": "One-use call approvals, signature-checked webhooks and a live question relay make the phone bridge a meaningful new agent capability.", "worst": "Real provider behavior is mocked in the suite, setup requires a public bridge sharing SQLite with Hermes, and free-text privacy filtering covers fewer patterns than the blanket consent claims suggest.", "standout": false, "blocked": null }, { "pr": 152771, "name": "assetrizz", "does": "Searches a bundled game-art catalog and composes, previews and exports layered 2D scenes using verified assets fetched on demand.", "useful": 3, "reliable": 4, "polish": 4, "creative": 4, "beyond": 4, "total": 23, "best": "Combines searchable art, deterministic scene composition, layout diagnostics and a Desktop pane with hash-verified, bounded downloads.", "worst": "The README and PR explicitly disclose that the Desktop pane has not been proven on a live Hermes Desktop host.", "standout": false, "blocked": null }, { "pr": 153395, "name": "hermes-guardian", "does": "Enforces deterministic tool and local-path policies through native approval hooks, with redacted audits and an operator policy simulation and activation workflow.", "useful": 3, "reliable": 4, "polish": 4, "creative": 3, "beyond": 5, "total": 24, "best": "Fail-closed policy parsing and audit failures, fresh approval keys and reviewed activation provide unusually careful enforcement boundaries.", "worst": "The terminal grammar permits only seven simple commands, and allowing an external tool checks its name without inspecting its arguments, sharply limiting practical protected workflows.", "standout": false, "blocked": null }, { "pr": 153787, "name": "pilot", "does": "Gives Hermes persistent interactive tmux terminals it can read and drive with text and keystrokes while a human can attach to the same pane.", "useful": 4, "reliable": 3, "polish": 4, "creative": 3, "beyond": 4, "total": 22, "best": "A small readable implementation and real-tmux tests cover persistent REPLs, vim keystrokes, screen polling and fast-exit recovery.", "worst": "All calls share the fixed hermes-pilot socket and ignore session context, so separate Hermes profiles or conversations can list and control each other's panes.", "standout": false, "blocked": null }, { "pr": 150309, "name": "privacy-shield", "does": "Pseudonymizes detected personal data in outgoing user messages and model reasoning, restores approved-channel replies and provides audit and Desktop controls.", "useful": 4, "reliable": 2, "polish": 4, "creative": 3, "beyond": 5, "total": 23, "best": "Request middleware, reasoning-intake hooks and session-scoped reversible tokens address privacy at several real Hermes lifecycle seams with broad regression coverage.", "worst": "The pinned code still reads PRIVACY_SHIELD_KEY_FILE outside Hermes secret scope, disables shielding when a key is absent, and documents unmasked core summary calls despite broad compliance claims.", "standout": false, "blocked": null }, { "pr": 152991, "name": "survey-desk", "does": "Computes and adjusts closed survey traverses, diagnoses misclosure, converts coordinate systems and exports annotated plans in image and geospatial formats.", "useful": 3, "reliable": 3, "polish": 4, "creative": 4, "beyond": 3, "total": 20, "best": "Independent area checks, ranked blunder candidates and explicit datum-shift accuracy make the survey output more useful than unchecked model arithmetic.", "worst": "Bearing parsing extracts unsigned numbers, so a string such as -5 is accepted as positive 5, and the polygon check misses collinear overlaps and endpoint touches.", "standout": false, "blocked": null }, { "pr": 151558, "name": "reset-rescue", "does": "Schedules redemption of expiring Codex reset credits across Hermes pool accounts, verifies results by readback and provides reminders and Desktop controls.", "useful": 3, "reliable": 3, "polish": 5, "creative": 3, "beyond": 4, "total": 22, "best": "Centralized consume authorization, persistent request IDs, readback confirmation and extensive failure-path tests support a polished account-wide rescue workflow.", "worst": "It depends on undocumented ChatGPT mutation endpoints and explicitly deviates from catalog credential routing by handling pool tokens directly; automatic irreversible redemption is enabled by default.", "standout": false, "blocked": null }, { "pr": 149421, "name": "aurora-design-system", "does": "Adds Aurora design-system skills, bundled component lookup and an offline HTML class and CSS-pack validator.", "useful": 3, "reliable": 2, "polish": 3, "creative": 2, "beyond": 3, "total": 16, "best": "Bundled canonical markup and census-backed class checks help agents avoid inventing design-system components without a service or API key.", "worst": "The validator only recognizes double-quoted lowercase class attributes and style blocks, misses equivalent HTML and inline CSS, and ships no tests; the README also incorrectly calls it skills-only.", "standout": false, "blocked": null } ] }