{
  "batch": "final-08",
  "ranking": [
    {
      "rank": 1,
      "pr": 152458,
      "name": "housemate",
      "total": 30,
      "why": "Useful 5, Reliable 5, Polish 5, Creative 5, Beyond a prompt 5: its paired-browser room platform, persistent timers, wake arbitration and device-bound approvals most fully turn the user's existing hardware into everyday Hermes access, supported by the local Hermes-aware suite, although live household audio was not remeasured here ([pinned evidence](https://raw.githubusercontent.com/RaYYeR220/housemate/09c1a1d0a88bfa8762cee6b3ccf94271b0f172c5/core/hub.py))."
    },
    {
      "rank": 2,
      "pr": 150840,
      "name": "kata",
      "total": 29,
      "why": "Useful 5, Reliable 5, Polish 5, Creative 4, Beyond a prompt 5: mining repeated successful tool sequences into owner-approved deterministic tools directly improves the agent's everyday efficiency, with rendered-argument risk checks, schema-drift retirement and a passing local suite giving it the strongest general-purpose implementation after Housemate ([pinned evidence](https://raw.githubusercontent.com/opeblow/kata/cc59b08f1bbeeb02524c060faac04edb34de0564/README.md))."
    },
    {
      "rank": 3,
      "pr": 153844,
      "name": "fine-print",
      "total": 29,
      "why": "Useful 5, Reliable 4, Polish 5, Creative 5, Beyond a prompt 5: automatic notices and model annotations for otherwise hidden context loss offer unusually broad value without extra model calls, but detectors tied to specific Hermes formats and reference-only coverage keep it just behind Kata ([pinned evidence](https://raw.githubusercontent.com/Joko-F/hermes-fine-print/7b574f1b5c8dc8eafe3ea9f87ae6b0111853f0ba/README.md))."
    },
    {
      "rank": 4,
      "pr": 153714,
      "name": "tenure",
      "total": 29,
      "why": "Useful 5, Reliable 4, Polish 5, Creative 5, Beyond a prompt 5: per-action revocable trust, authenticated undo and a signed evidence ledger make oversight a substantial native agent capability with excellent UI polish, while conservative suspension after external content and classification complexity reduce everyday ease and reliability ([pinned evidence](https://raw.githubusercontent.com/otsoa-dubief/tenure/f4a62855e592b9e170ab0c54d9278aabef05f370/README.md))."
    },
    {
      "rank": 5,
      "pr": 149640,
      "name": "hermes-ark",
      "total": 28,
      "why": "Useful 5, Reliable 5, Polish 5, Creative 3, Beyond a prompt 5: automatic pre-write snapshots, encryption, credential exclusion and granular recoverable restores address almost every long-lived Hermes installation, and its passing backup/security suite and low daily friction win the 28-point retention tie despite a familiar underlying backup idea ([pinned evidence](https://raw.githubusercontent.com/medomohamed22/hermes-ark/a23e529c1f636aa49d974576cdc5cb15fc8242d3/README.md))."
    },
    {
      "rank": 6,
      "pr": 150556,
      "name": "hermes-ui-cards",
      "total": 28,
      "why": "Useful 5, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: validated interactive cards and persistent answer state improve many Desktop conversations while preserving text fallbacks and core approvals, but eventual cross-client answer locking and model-mediated stale confirmations stop the polished interaction layer earning full reliability ([pinned evidence](https://raw.githubusercontent.com/robbyczgw-cla/hermes-ui-cards/c40df4d41cd5439cf4902995fd4a9e5505483f25/README.md))."
    },
    {
      "rank": 7,
      "pr": 153955,
      "name": "shadow",
      "total": 28,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 5, Beyond a prompt 5: blind paired replays, isolated tool handling and reversible model switching directly improve the user's own model choice, but required live isolation verification, private-helper coupling and paid comparisons not exercised here make it less immediately useful than the entries above ([pinned evidence](https://raw.githubusercontent.com/akinduroifedayo/shadow/38173bc9c696df3b1f87b5b9b60ccecd36a260ab/README.md))."
    },
    {
      "rank": 8,
      "pr": 150800,
      "name": "hypnos",
      "total": 28,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 5, Beyond a prompt 5: real SFT/DPO training, held-out promotion exams and native routing make procedural habits becoming model weights one of the most original entries, but the separate trainer and unverified live training path in this review narrow its day-one audience ([pinned evidence](https://raw.githubusercontent.com/0xNoramiya/hypnos/0a809b9085fdeaed63fd909b927afcc41de512de/README.md))."
    },
    {
      "rank": 9,
      "pr": 153020,
      "name": "breadcrumbs",
      "total": 28,
      "why": "Useful 5, Reliable 4, Polish 4, Creative 5, Beyond a prompt 5: exact-source memory zoom, deterministic retrieval and a cache-conscious context engine provide a distinctive daily foundation that worked when activated in Hermes, although provider/context configuration and the experimental breadth make it less finished than the leading products ([pinned evidence](https://raw.githubusercontent.com/breadcrumbs-agency/hermes-breadcrumb-plugin/b8975645e5174edbcc00d5dc86a78e536bce938b/README.md))."
    },
    {
      "rank": 10,
      "pr": 150596,
      "name": "ariadne",
      "total": 28,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 5, Beyond a prompt 5: recording and replaying the real Hermes loop without executing recorded tools is valuable native debugging infrastructure, but a same-build CLI recording replayed through the TUI reported platform/tool drift locally, weakening the upgrade-check promise and keeping this below broader daily memory tools ([pinned evidence](https://raw.githubusercontent.com/miracle352/hermes-ariadne/128c95967a0761ad24d3d224441d1502c43bbb69/driver.py))."
    },
    {
      "rank": 11,
      "pr": 149738,
      "name": "trust-ladder",
      "total": 28,
      "why": "Useful 5, Reliable 4, Polish 4, Creative 5, Beyond a prompt 5: turning corrections into example-checked runtime guards has broad retention value and really blocked a learned forbidden push in Hermes, while best-effort pattern coverage and fail-open hook errors limit its safety guarantee and its interface is less finished than the leaders ([pinned evidence](https://raw.githubusercontent.com/merdoks/trust-ladder/fae4e9df58767e5e3832c8f65936b97868bc98fc/README.md))."
    },
    {
      "rank": 12,
      "pr": 153851,
      "name": "foreman",
      "total": 28,
      "why": "Useful 5, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: deterministic anti-rigging checks, exact-diff secondary review and owner-controlled merging form a strong coding product, but disclosed shell/API escape routes and advisory model coverage make its narrower integrity guarantee less compelling than general correction enforcement ([pinned evidence](https://raw.githubusercontent.com/PedValme/foreman/c15453cb2fc07fa674d6f7b548a553b3ce9e6f5d/plugin/README.md))."
    },
    {
      "rank": 13,
      "pr": 150353,
      "name": "fluent",
      "total": 28,
      "why": "Useful 5, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: learning installed CLI help and blocking invented options is a useful everyday extension that preserves Hermes approvals, but local-backend and high-confidence/manual coverage restrictions leave gaps in the very commands it aims to protect ([pinned evidence](https://raw.githubusercontent.com/Bruno-coder21/fluent/182c4301f2437e784e9022f1f38576e1d68c3f5f/README.md))."
    },
    {
      "rank": 14,
      "pr": 152691,
      "name": "grimoire",
      "total": 28,
      "why": "Useful 4, Reliable 5, Polish 5, Creative 4, Beyond a prompt 5: read-only constrained DuckDB queries, numeric provenance checks and a finished Desktop data explorer make an unusually dependable specialist, with local aggregation and write rejection working in Hermes, but fewer users would retain it than the broader terminal guard above ([pinned evidence](https://raw.githubusercontent.com/bettersandeep/grimoire/4824773040b131f3179e5896d71d1435908758a7/engine.py))."
    },
    {
      "rank": 15,
      "pr": 151887,
      "name": "cmi",
      "total": 27,
      "why": "Useful 5, Reliable 4, Polish 4, Creative 4, Beyond a prompt 5: deterministic cache-aware context folding and exact-output recall offer broad token-saving potential and its engine/search/error paths worked when activated in Hermes, but private compressor/pricing dependencies and a less finished native user experience keep it below the more complete products ([pinned evidence](https://raw.githubusercontent.com/jeremiebergel/context-manager-intelligence/399468f0687c051476961e4bb045eb84263be6b1/cmi/engine.py))."
    },
    {
      "rank": 16,
      "pr": 150644,
      "name": "hermes-canvas",
      "total": 27,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: a polished plan that reacts to real file/check events and accepts mid-build user vetoes adds genuine Desktop control, with create/read/error paths working locally, but heuristic shell attribution and coding-specific usefulness put it behind more universal context improvements ([pinned evidence](https://raw.githubusercontent.com/0xbyt4/hermes-canvas/d6eeff0d6f7ec653262537bdd7bb001cc6d3c7f8/README.md))."
    },
    {
      "rank": 17,
      "pr": 150251,
      "name": "gemini-live",
      "total": 27,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: full-duplex voice delegation, thread steering and locked ephemeral credentials add a substantial Desktop/dashboard interface through proper secret routes, but an external Gemini account, optional companion picker integration and live calls not exercised here reduce its reach and confidence ([pinned evidence](https://raw.githubusercontent.com/cineglobe/hermes-gemini-live/b6c2f24354aeed1a8b5651b62944d14ed96a472d/glive_core.py))."
    },
    {
      "rank": 18,
      "pr": 152908,
      "name": "contract-necromancer",
      "total": 27,
      "why": "Useful 4, Reliable 4, Polish 5, Creative 4, Beyond a prompt 5: owner-frozen tests, fresh verification and recheckable completion receipts enforce more than a coding prompt can, but bounded nudges, short gate budgets and shell-only edits being stamped rather than kept in the repair loop make it less complete than the stronger coding-control entries ([pinned evidence](https://raw.githubusercontent.com/vijaygopalbalasa/hermes-necromancer/0ff7bec40aba10091a84f94025d4471984fd7386/README.md))."
    },
    {
      "rank": 19,
      "pr": 153809,
      "name": "living-spaces",
      "total": 25,
      "why": "Useful 4, Reliable 3, Polish 4, Creative 4, Beyond a prompt 5: persistent sandboxed mini-apps with shared records and revision recovery worked through the native tools, but private cron modules, a raw Desktop bridge and macOS-only qualification weaken compatibility and finish despite substantial platform ambition ([pinned evidence](https://raw.githubusercontent.com/akinduroifedayo/living-spaces/e6145ccbd028edf55417822020426f4830343eb6/src/living-spaces/README.md))."
    },
    {
      "rank": 20,
      "pr": 152651,
      "name": "workroom",
      "total": 25,
      "why": "Useful 4, Reliable 1, Polish 5, Creative 5, Beyond a prompt 5: its ambitious Office provenance and owner-approved release engine is undermined by the pinned tripwire automatically reading project .env/key-file bytes directly, which I treat as a catalogue credential-rule blocker despite its disclosed local-only handling ([pinned evidence](https://raw.githubusercontent.com/Pratiikpy/workroom-for-hermes/c5f6bf334c1163951e203dc7c5777faadfe49795/README.md))."
    }
  ]
}