{ "batch": "read1-021", "plugins": [ { "pr": 0, "name": "done-bell", "does": "Notifies on completed turns and approval waits through OS alerts, chimes and a Desktop status chip.", "useful": 4, "reliable": 3, "polish": 3, "creative": 2, "beyond": 4, "total": 20, "best": "Real lifecycle hooks and a Desktop watcher provide useful notifications without credentials or an external service.", "worst": "Installation examples are developer-specific Windows paths, Desktop behavior lacks tests, and notification failures are silently swallowed.", "standout": false, "blocked": null }, { "pr": 0, "name": "homeassistant", "does": "Connects Hermes to Home Assistant device tools, state-change events and notification delivery.", "useful": 4, "reliable": 5, "polish": 4, "creative": 3, "beyond": 5, "total": 26, "best": "A full gateway platform combines scoped secrets, bounded requests, filtered event delivery, service restrictions and focused adapter/policy tests.", "worst": "Its safety boundary still depends on the Home Assistant token and optional entity filters; scripts and automations can have indirect effects.", "standout": false, "blocked": null }, { "pr": 153882, "name": "claim-check", "does": "Annotates agent success claims with deterministic receipts derived from a local tool-result ledger.", "useful": 4, "reliable": 3, "polish": 3, "creative": 4, "beyond": 4, "total": 22, "best": "Observes actual tool results through hooks and tests contradictions, stale tests, session separation and optional correction nudges.", "worst": "Regex matching can verify quoted or compound commands without proving execution, shell edits do not stale test receipts, and stored command/output fragments have no secret redaction.", "standout": false, "blocked": null }, { "pr": 154036, "name": "retake", "does": "Provides a Desktop workspace for image generation, annotated revision briefs, immutable takes and side-by-side review.", "useful": 4, "reliable": 4, "polish": 5, "creative": 3, "beyond": 5, "total": 26, "best": "Durable single-use generation reservations, profile checks, bounded media handling and extensive boundary tests make paid image iteration unusually careful.", "worst": "Generation requires extra Hermes tool-server setup and disabling tool search for the profile; the exact URL adapter and final UI lack a claimed full live end-to-end pass.", "standout": false, "blocked": null }, { "pr": 151917, "name": "aegis-spend-guard", "does": "Applies owner-defined spending caps, recipient policies and a freeze switch before spend-shaped tool calls.", "useful": 4, "reliable": 3, "polish": 3, "creative": 3, "beyond": 4, "total": 21, "best": "Uses real pre-tool block/approval directives and a tested integer-cent policy engine rather than asking the model to obey a budget.", "worst": "Detection is heuristic and the ledger has no atomic reservation for concurrent spends, so parallel calls can exceed the daily cap and dynamically constructed payments can evade detection.", "standout": false, "blocked": null }, { "pr": 149657, "name": "tiled-sessions", "does": "Adds keyboard-controlled session tiling, multiple layouts and saved workspaces to Hermes Desktop.", "useful": 4, "reliable": 3, "polish": 5, "creative": 3, "beyond": 5, "total": 25, "best": "A substantial pure layout engine, geometry tests, configurable shortcuts and illustrated UI make multi-session work much easier.", "worst": "Its defining tiling features depend on the proposed upstream session-layout SDK; without that seam it falls back to ordinary tabs, with no manifest release floor.", "standout": false, "blocked": null }, { "pr": 152387, "name": "hermes-psfn-avatar", "does": "Adds a reactive 3D VRM avatar, motion studio and optional PSFN emotion synchronization to Desktop and Dashboard.", "useful": 3, "reliable": 3, "polish": 4, "creative": 3, "beyond": 5, "total": 23, "best": "Real skeletal motion playback, mood envelopes, avatar uploads and lifecycle integration go well beyond a decorative prompt.", "worst": "The authenticated PSFN bridge validates only the initial loopback URL and uses default redirect-following urllib, leaving a potential bearer-token leak through a redirected endpoint.", "standout": false, "blocked": null }, { "pr": 151920, "name": "explainer", "does": "Renders model-authored storyboards into narrated MP4 explainers with synchronized captions, slides and a local video library.", "useful": 4, "reliable": 4, "polish": 4, "creative": 3, "beyond": 3, "total": 21, "best": "Implements timed narration, caption alignment, video encoding and tested SSRF-resistant image retrieval with actionable setup diagnostics.", "worst": "Narration depends on an online Edge TTS service and local ffmpeg, and the presentation workflow largely remains conventional tooling that a skill could orchestrate.", "standout": false, "blocked": null }, { "pr": 153465, "name": "threejs-stage", "does": "Shows generated 3D scenes and local models in interactive chat cards, with hidden runtime checks and repair feedback.", "useful": 4, "reliable": 3, "polish": 5, "creative": 4, "beyond": 5, "total": 26, "best": "Integrates sandboxed live rendering, model import, GPU dry-run feedback and repair controls directly into Hermes chat.", "worst": "No automated behavioral test suite is shipped for its sizable tool/backend; exported HTML runs model-written code without the Desktop sandbox CSP and loads a CDN at runtime.", "standout": false, "blocked": null }, { "pr": 151630, "name": "hermes-blueprint", "does": "Turns ideas into evidence-backed plans through stored phase gates, independent research checks, estimates and kanban handoff.", "useful": 4, "reliable": 4, "polish": 4, "creative": 4, "beyond": 4, "total": 24, "best": "Quote-checked evidence, persisted verdicts, computed effort/calendar estimates and focused offline tests enforce useful planning discipline.", "worst": "A failed or malformed independent verifier is stored as skipped and still satisfies the briefing freshness gate, so a plan can advance without a successful independent check.", "standout": false, "blocked": null }, { "pr": 151301, "name": "confidential-mode", "does": "Adds profile privacy tiers, attested encrypted inference transports, off-host tool approval gates and replayable evidence receipts.", "useful": 4, "reliable": 4, "polish": 4, "creative": 5, "beyond": 5, "total": 27, "best": "Implements real model-execution middleware, TLS/attestation pinning, encrypted transports and extensive hostile, fault and cryptographic-vector tests.", "worst": "The lock cannot stop auxiliary calls that drift mid-turn or network access from shell/code tools, and its SDK/private-core dependencies make this a bounded privacy guard rather than complete isolation.", "standout": true, "blocked": null }, { "pr": 150160, "name": "helm", "does": "Adds a Desktop control desk for bot activity, pending approvals and questions, live plans, themes and robot avatars.", "useful": 5, "reliable": 4, "polish": 5, "creative": 3, "beyond": 5, "total": 27, "best": "Combines a finished multi-bot interface with click-driven approval handling, live task plans, detailed guides and SDK-mock/router tests.", "worst": "Cross-gateway inbox coverage is limited, CSS depends on core slot names, and approval commands are truncated to 160 characters in the desk row.", "standout": true, "blocked": null } ] }