{"batch":"read1-130","plugins":[{"pr":0,"name":"done-bell","does":"Notifies you when a Hermes turn finishes or waits for approval, with an optional desktop status chip and chime.","useful":3,"reliable":4,"polish":4,"creative":3,"beyond":3,"total":20,"best":"It is a small, practical desktop notification feature with clear hooks, mute controls, and no credentials.","worst":"Its value is narrow and the Windows-oriented installation instructions are awkward for a cross-platform catalog plugin.","standout":false,"blocked":null},{"pr":0,"name":"homeassistant","does":"Connects Hermes to Home Assistant events and exposes filtered entity/state/service tools for smart-home control.","useful":5,"reliable":4,"polish":5,"creative":3,"beyond":5,"total":27,"best":"It turns Hermes into a real smart-home platform while adding strong entity filtering and service safety checks.","worst":"Control still depends on a broad Home Assistant token and the adapter is a substantial network integration whose safety cannot fully exceed that token.","standout":true,"blocked":null},{"pr":153443,"name":"local-hands","does":"Runs Hermes on an always-on server while routing terminal and file operations to an approved local computer through a pinned-TLS daemon.","useful":5,"reliable":4,"polish":4,"creative":5,"beyond":5,"total":28,"best":"It solves a major deployment limitation with an unusually complete pairing, approval, audit, reconnect, and multi-machine design.","worst":"The security-critical custom protocol and detached hub are complex, so operational setup and review remain demanding despite the end-to-end tests.","standout":true,"blocked":null},{"pr":149823,"name":"samskara","does":"Provides relational, multi-hop graph memory with provenance, weighting, and explicit conflict detection instead of similarity-only recall.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"best":"It changes Hermes memory semantics in a genuinely agent-specific way and backs the design with extensive offline tests, demos, and provenance.","worst":"The graph backend and optional model extraction add deployment complexity, and the snapshot has relatively few conventional test files compared with its advertised coverage.","standout":true,"blocked":null},{"pr":149360,"name":"harness2","does":"Runs isolated probe sessions to propose, test, compose, and adopt evidence-backed changes to an agent profile’s prompt, config, and skills.","useful":4,"reliable":3,"polish":4,"creative":5,"beyond":5,"total":26,"best":"It makes self-improvement an actual measured loop with isolated profiles, budgets, structural gates, trajectories, and a reusable library.","worst":"There are no repository tests, and the central subprocess/profile orchestration has many failure and environment assumptions.","standout":true,"blocked":null},{"pr":151680,"name":"hermes-wrapped","does":"Turns Hermes session records into a deterministic personalized weekly or periodic recap video with synthesized music and optional delivery.","useful":3,"reliable":4,"polish":5,"creative":5,"beyond":4,"total":25,"best":"It is an unusually polished, privacy-aware product experience with deterministic analytics, themed visuals, music, screenshots, and tests.","worst":"It is mostly a delightful retrospective rather than a capability that improves everyday agent work, and rendering requires large external tooling.","standout":true,"blocked":null},{"pr":151594,"name":"hermes-x-bookmarks","does":"Syncs X bookmarks through the official API, classifies them with Hermes, and exports them to Desktop, Obsidian, Notion, or Google Docs.","useful":4,"reliable":3,"polish":5,"creative":4,"beyond":4,"total":24,"best":"It connects a real personal knowledge workflow across several useful destinations with OAuth scopes, a Desktop board, digest actions, and a substantial test suite.","worst":"The number of auth and export integrations creates many configuration and API failure modes, while the core classification remains model-dependent.","standout":false,"blocked":null},{"pr":149231,"name":"proofchain","does":"Records Hermes tool calls in a redacted hash chain with HMAC sealing, verification, search, and signed export for audits.","useful":3,"reliable":4,"polish":4,"creative":4,"beyond":4,"total":23,"best":"The durable tamper-evident ledger is a concrete audit primitive with secret redaction and verification surfaces integrated at tool boundaries.","worst":"It largely records activity rather than improving the agent’s work, and its HMAC key lifecycle and evidence semantics deserve more external operational documentation.","standout":false,"blocked":null},{"pr":150891,"name":"secretshield","does":"Scans tool arguments, tool results, and final replies for credential material and blocks or redacts leaks while keeping a metadata-only incident ledger.","useful":5,"reliable":4,"polish":4,"creative":4,"beyond":5,"total":27,"best":"It addresses a high-value agent-specific failure mode at the exact pre-dispatch and pre-delivery boundaries, with fail-closed behavior and tests for secret non-persistence.","worst":"Pattern detection cannot guarantee semantic secret discovery, and blocking or rewriting every surface can create compatibility and false-positive costs.","standout":true,"blocked":null},{"pr":152207,"name":"codecoach","does":"Builds small C or Python learning projects step by step, checking student code in a sandbox with visible and hidden tests and constrained hints.","useful":5,"reliable":4,"polish":5,"creative":4,"beyond":5,"total":28,"best":"It turns Hermes into a real interactive programming tutor whose examples, progression, tests, hints, and Desktop UI are enforced by code.","worst":"Lesson quality and much of the project plan remain model-dependent, while sandbox, compiler, dependency, and long local-model setup requirements are substantial.","standout":true,"blocked":null},{"pr":153350,"name":"freebusy","does":"Schedules meetings by email from free/busy data, holding offered slots, enforcing timezone and approval policy, and booking only through guarded paths.","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":4,"total":27,"best":"Its invariant-driven policy, hold ledger, timezone handling, owner approvals, reminders, fixture demo, and broad tests make a risky automation domain unusually careful.","worst":"The pinned release explicitly leaves the Google Workspace adapter unwired, so the impressive tested fixture path does not yet deliver the advertised real-calendar integration.","standout":true,"blocked":null},{"pr":149490,"name":"screenpipe","does":"Searches a local Screenpipe OCR/audio corpus and injects relevant timestamped context before Hermes sees a turn, with explicit recall tools.","useful":4,"reliable":4,"polish":4,"creative":4,"beyond":4,"total":24,"best":"The pre-LLM hook, relevance gating, cooldown cache, untrusted-context prefix, bounded timeout, and graceful degradation make ambient recall practical.","worst":"It requires a separate always-on Screenpipe capture service and sends highly sensitive local context into the model, even though the plugin itself is careful about limits.","standout":false,"blocked":null}]}