{ "batch": "read1-091", "plugins": [ { "pr": 0, "name": "done-bell", "does": "Notifies you when a Hermes turn finishes or needs approval, with desktop status, OS notifications, and an optional chime.", "useful": 3, "reliable": 3, "polish": 4, "creative": 2, "beyond": 3, "total": 18, "best": "It connects session and approval hooks to an attractive Desktop status chip and notifications without needing credentials.", "worst": "Its central feature is a narrow notification convenience, and the test surface is small compared with the more involved plugins.", "standout": false, "blocked": null }, { "pr": 0, "name": "homeassistant", "does": "Connects Hermes to Home Assistant as a gateway platform and exposes tools to list, inspect, and control smart-home entities.", "useful": 5, "reliable": 4, "polish": 5, "creative": 3, "beyond": 5, "total": 27, "best": "It combines real two-way platform integration with useful device tools, a substantial test suite, clear setup, and entity allow/deny controls.", "worst": "Device control still depends on a user-supplied long-lived Home Assistant token, and broad automations require careful local configuration.", "standout": true, "blocked": null }, { "pr": 151450, "name": "rehearse", "does": "Turns Hermes Desktop into an interview and presentation practice room with spoken questions, local speech and face analysis, scored replays, and model-written coaching.", "useful": 4, "reliable": 4, "polish": 5, "creative": 4, "beyond": 5, "total": 27, "best": "The integrated Desktop workflow combines live voice, local recording analysis, detailed replay, and tailored coaching, with seven focused test modules and clear privacy disclosure.", "worst": "The experience depends on sizable platform-specific speech and vision dependencies, and Windows ARM loses full analysis.", "standout": true, "blocked": null }, { "pr": 153531, "name": "hermes-reliability", "does": "Builds evidence-backed agent reliability campaigns from source snapshots and datasets, then reviews, executes, and freezes them for regression comparisons.", "useful": 4, "reliable": 3, "polish": 5, "creative": 5, "beyond": 5, "total": 27, "best": "A rich end-to-end testing workbench joins scenario generation, independent review, evidence audits, frozen benchmarks, Hermes memory references, and dashboard views.", "worst": "The breadth creates a large compatibility and execution surface: generated evaluators can run on the host, parts of the target harness use private APIs, and full autonomous live campaigns remain unverified.", "standout": true, "blocked": null }, { "pr": 149267, "name": "escalation-governor", "does": "Detects repeated tool failures and stalled model progress, then adds an escalation packet directing Hermes to delegate or report the blocker.", "useful": 4, "reliable": 3, "polish": 3, "creative": 4, "beyond": 5, "total": 24, "best": "Deterministic, configurable signals can nudge Hermes out of retry loops without adding a network service or blocking tool calls.", "worst": "The packet only directs the model to delegate; the plugin does not enforce that escalation, and its sole unit-test file offers limited assurance for the many lifecycle hooks.", "standout": false, "blocked": null }, { "pr": 153289, "name": "the-millionaire-matrix", "does": "Runs competing model-generated code candidates through test and measured gates, blind peer scoring, a tournament, and a human approval gate, then records lessons for later sessions.", "useful": 4, "reliable": 3, "polish": 4, "creative": 5, "beyond": 5, "total": 26, "best": "The eight-family competition, hard evidence gates, explicit human merge approval, route holds for degraded judging, and broad test suite make an unusually ambitious agent coding workflow.", "worst": "It executes operator-authored commands and generated candidates in the target worktree, while extensive adapters and runtime assumptions increase operational and maintenance risk.", "standout": true, "blocked": null }, { "pr": 152051, "name": "fullenrich", "does": "Uses a hosted MCP service and guided workflows to search B2B contacts, enrich verified details, and export results.", "useful": 3, "reliable": 3, "polish": 4, "creative": 2, "beyond": 3, "total": 18, "best": "It offers a concrete enrichment service with OAuth managed by Hermes and repeatedly calls out that paid operations consume account credits.", "worst": "The pinned package is primarily instructional skills and remote connection metadata; the actual MCP implementation is hosted, and end-to-end OAuth and paid workflows were not tested.", "standout": false, "blocked": null }, { "pr": 151946, "name": "stitchslop", "does": "Lets Hermes inspect and operate the Stitch Slop embroidery digitizer through a local bridge, a live Chrome pane, digitizing tools, and DST/PES export.", "useful": 3, "reliable": 4, "polish": 5, "creative": 5, "beyond": 5, "total": 27, "best": "It makes a real graphical specialist app controllable and observable by the agent while keeping the person in the loop, with carefully gated pairing and substantive security tests.", "worst": "The use case is highly specialized and Linux-only, and it requires a separately running browser, X display, and manual first-time token pairing.", "standout": true, "blocked": null }, { "pr": 152349, "name": "tldr", "does": "Adds a Desktop control that summarizes a selected span of the current chat into status, evidence labels, risks, and decisions users can review and insert.", "useful": 4, "reliable": 4, "polish": 5, "creative": 3, "beyond": 5, "total": 26, "best": "It reads actual session history and returns structured, revisable Desktop cards without consuming a chat turn, with user-controlled scope and opt-in capped automation.", "worst": "Two private-core imports and the complex session-history projection create compatibility risk, and the PR reports no project-specific automated test count.", "standout": true, "blocked": null }, { "pr": 153728, "name": "plugin-safehouse", "does": "Trials a local Hermes plugin in a disposable macOS sandbox with planted canaries, invokes reachable handlers, and reports observed behavior and coverage.", "useful": 4, "reliable": 4, "polish": 4, "creative": 5, "beyond": 5, "total": 27, "best": "Behavioral execution with fake secrets, a jailed worker, handler coverage, and careful INCONCLUSIVE/REVIEW distinctions addresses a real gap left by static scanning; it includes extensive attack fixtures and tests.", "worst": "Dynamic testing is macOS-only and supports only standalone and exclusive plugins; its real-loader integration relies on a disclosed private API seam.", "standout": true, "blocked": null }, { "pr": 153291, "name": "hermes-gauge", "does": "Tracks estimated context-window use by session, reports the effective window and token breakdown, and can alert when a session approaches configured thresholds.", "useful": 4, "reliable": 3, "polish": 4, "creative": 3, "beyond": 5, "total": 24, "best": "It turns session lifecycle and model-window data into actionable per-session readings and opt-in alerts, with histories, dashboard surfaces, and screenshots.", "worst": "Token counts are approximate and can be UNKNOWN until enough information exists; much of the live validation is described but the PR contains no test or check results.", "standout": false, "blocked": null }, { "pr": 152248, "name": "rainbowomma", "does": "Separates each Hermes profile’s browser and leased resources on macOS and guards tool calls that could disrupt the human or another agent.", "useful": 4, "reliable": 3, "polish": 5, "creative": 5, "beyond": 5, "total": 27, "best": "The ownership registry, dedicated browser provider, leases, approval escalation, pre-tool guard, and cleanup solve cross-agent and human-resource collisions at the operating-system boundary.", "worst": "It is a macOS-only system integration with many powerful process, document, and GUI actions; its guard is explicitly not a sandbox and its project exposes only two test modules.", "standout": true, "blocked": null } ] }