{"batch":"read1-060","plugins":[{"pr":0,"name":"done-bell","does":"Rings and shows status when a Hermes turn finishes or needs approval, with optional desktop controls.","useful":4,"reliable":4,"polish":4,"creative":3,"beyond":4,"total":23,"best":"It is a small, coherent cross-surface quality-of-life plugin with persistent state and defensive error handling.","worst":"The desktop and notification paths are platform-specific and the tests cover only part of the integration surface.","standout":false,"blocked":null},{"pr":0,"name":"homeassistant","does":"Connects Hermes to Home Assistant events, notifications, cron delivery, and REST tools for inspecting and controlling devices.","useful":5,"reliable":5,"polish":5,"creative":4,"beyond":5,"total":29,"best":"It changes where Hermes runs and what real household hardware it can observe and control, with filters, cooldowns, secret-store configuration, and a mature test suite.","worst":"The integration necessarily depends on a live Home Assistant instance and its external API semantics.","standout":false,"blocked":null},{"pr":150379,"name":"gateway-guardian","does":"Reads Hermes gateway logs and processes to classify Telegram health and produce an operator-approved recovery plan.","useful":4,"reliable":4,"polish":4,"creative":3,"beyond":4,"total":23,"best":"The read-only, approval-gated design makes a vague gateway failure actionable without silently restarting or killing processes.","worst":"It is a narrow pattern scanner with no hooks or live remediation, so detection quality depends on local log/process conventions.","standout":true,"blocked":null},{"pr":150678,"name":"precedent-log","does":"Stores board-scoped human rulings, injects applicable context for Kanban workers, and can veto repeated questions while exposing CLI and Desktop review.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"best":"It turns prior human decisions into durable, scoped agent context with unusually strong audit, Desktop, recovery, and safety disclosures.","worst":"The cooperative veto can leave a worker card running, and matching text is not proof that the ruling was actually applied.","standout":true,"blocked":null},{"pr":153970,"name":"defuzz","does":"Audits Hermes homes and organizes projects with templates, AGENTS.md rules, placement hooks, linting, and Desktop/dashboard surfaces.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"best":"It materially improves project hygiene and agent write placement across CLI, hooks, Desktop, and dashboard, with clear disclosure of its broad scope.","worst":"Its breadth and optional code/lint execution create a large trust and compatibility surface, and it is Linux/Python-3.13-only.","standout":true,"blocked":null},{"pr":153480,"name":"household-fleet","does":"Applies a deterministic message gate with caps, quiet hours, deduplication and audit history, and mirrors tasks into Home Assistant.","useful":5,"reliable":4,"polish":4,"creative":4,"beyond":5,"total":27,"best":"The state-hash anti-repeat policy and pure, offline-testable decision function address a real multi-agent failure mode mechanically.","worst":"The plugin returns a delivery decision rather than delivering, and the Home Assistant mirror adds a live-service dependency.","standout":true,"blocked":null},{"pr":151043,"name":"hermes-token-ledger","does":"Records token counts from API-response hooks in SQLite and reports estimated model cost with optional budget warnings.","useful":4,"reliable":2,"polish":3,"creative":2,"beyond":4,"total":19,"best":"It uses a core lifecycle hook to make session-level usage and cost visible instead of relying on provider dashboards.","worst":"The pinned repository has no tests and a small implementation, while its pricing table and token shapes can become stale or incomplete.","standout":false,"blocked":null},{"pr":154045,"name":"wake","does":"Queues notes and cron reports, waits for the bound chat, optionally prechecks relevance, and injects a turn into the user’s own agent conversation.","useful":5,"reliable":4,"polish":4,"creative":4,"beyond":5,"total":27,"best":"It adds a genuinely Hermes-native proactive-agent loop across cron, platform delivery, CLI, hooks, and injected turns while defaulting to quiet skips.","worst":"It relies on a polling queue and gateway/session behavior, and retries can complicate operational diagnosis when a chat is unavailable.","standout":true,"blocked":null},{"pr":153684,"name":"postkey","does":"Connects Hermes to Postkey’s hosted MCP service for Instagram messaging, automations, contacts, and scheduled posts.","useful":4,"reliable":2,"polish":4,"creative":3,"beyond":3,"total":19,"best":"The package clearly discloses approvals, untrusted-server handling, OAuth, remote effects, and the fact that the real tool set is not pinned locally.","worst":"The pinned entry contains no executable plugin code and its 57-tool behavior is supplied by a mutable remote service, so reliability and claims are outside the commit.","standout":false,"blocked":null},{"pr":153134,"name":"roster-compiler","does":"Parses hospital rosters, reports uncovered shifts, double bookings, excessive runs and unsafe night-to-day transitions, then exports corrected CSV and HTML reports.","useful":4,"reliable":5,"polish":4,"creative":3,"beyond":3,"total":22,"best":"Deterministic parsers, explicit warnings instead of guesses, input protection, and focused tests make the safety-critical path trustworthy.","worst":"It is a specialized file tool without Hermes hooks, Desktop, or broader agent-platform changes, and PDF/image support has format limits.","standout":true,"blocked":null},{"pr":151227,"name":"perf-lab","does":"Benchmarks and profiles agent-edited programs, compares output fingerprints, and hooks Hermes to block unmeasured optimization claims.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"best":"It makes performance claims an evidence-producing Hermes workflow with output equivalence, noise handling, hooks, and a polished Desktop view.","worst":"The hooks fail open and scripts still execute with the user’s permissions, so it is measurement discipline rather than a security sandbox.","standout":true,"blocked":null},{"pr":152692,"name":"hermes-parallel","does":"Creates three isolated file futures, verifies and previews their diffs, and permits only an explicit human slash command to apply one.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"best":"The bounded operations, hash rechecks, fail-closed proof gate, and human-only apply path make multi-outcome agent editing concrete and inspectable.","worst":"Its filesystem isolation is not an OS security boundary, and model-backed planning is unverified unless the operator opts in.","standout":true,"blocked":null}]}