{
 "batch": "trial-3",
 "setup": "Hermes installed from a fresh clone of NousResearch/hermes-agent with uv sync --locked --python 3.14 (uv from astral.sh); tryit.py from the gist. Validate, enable and doctor ran for all six. The chat step (the scripted model) failed for all: the sandbox refuses the gateway control socket, so no planned calls ran inside Hermes; talos and folio were also called/tested directly, others judged by reading. Tests run with the Hermes venv plus the plugins' declared dependencies.",
 "plugins": [
  {
   "pr": 150732,
   "name": "folio",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed; doctor found 7 tools registered (folio_search, open, verify, changes, bibliography, index, forget), plus /folio command, dashboard API, desktop page and a skill. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "passed: 40 passed in its own pytest suite (after installing its declared deps openpyxl, pypdf, pypdfium2).",
   "useful": 4,
   "reliable": 4,
   "polish": 5,
   "creative": 4,
   "beyond": 4,
   "total": 25,
   "best": "A local, cited document library with quote verification and revision history, no network, with tests, a demo video and a desktop page.",
   "worst": "Its tools were not exercised through a model here, and it runs a background re-scan thread over user folders.",
   "blocked": null
  },
  {
   "pr": 152870,
   "name": "model-explainer",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed; registers nothing agent-side (no tools or hooks). All work is a Desktop page (desktop/plugin.js) with a bundled models.json, which could not be run here. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "none (a node data-validation script and CI check exist; not run).",
   "useful": 3,
   "reliable": 3,
   "polish": 4,
   "creative": 3,
   "beyond": 4,
   "total": 21,
   "best": "A sourced Desktop model picker companion with compare, a cost/score frontier, and a Use this model switch that only acts on a click; screenshots and a clear disclosure table.",
   "worst": "No tests, the Desktop UI was not run, and its model data is a hand-curated snapshot that fetches updates from the author's GitHub main branch.",
   "blocked": null
  },
  {
   "pr": 151264,
   "name": "toolfound",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed; a portable Agent Plugins package (plugin.json, mcp.json, skills) with no code of its own. Tools come at run time from the hosted MCP server at toolfound.com, so none could be seen or called. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "none (no code).",
   "useful": 2,
   "reliable": 3,
   "polish": 2,
   "creative": 1,
   "beyond": 3,
   "total": 14,
   "best": "Marked the remote server untrusted and discloses that every call goes to toolfound.com; submission needs user-confirmed fields.",
   "worst": "It is a thin wrapper for one vendor's product-directory service with skills that promote it; its tool list is not covered by the pinned commit and could not be checked.",
   "blocked": null
  },
  {
   "pr": 152847,
   "name": "waifu-sprites",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed; registers 6 lifecycle hooks (pre/post tool, pre/post llm, session start/end) and a dashboard tab with a FastAPI backend and overlay; no agent tools. Dashboard not run here. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "none.",
   "useful": 2,
   "reliable": 3,
   "polish": 4,
   "creative": 3,
   "beyond": 3,
   "total": 18,
   "best": "Polished dashboard companion that reacts to agent state, with screenshots, a docs page, and clips pinned by commit on a CDN.",
   "worst": "No tests; it is cosmetic, writes state.json beside the plugin code, and loads video from a third-party CDN.",
   "blocked": null
  },
  {
   "pr": 150921,
   "name": "talos",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed (1 tool, 3 hooks). Called directly: talos_scan on its fixtures found 9 findings (critical missing-signer, high unchecked-account etc.); on a nonexistent path it returned 0 findings with no error; the gate blocked 'anchor deploy ... mainnet' with the findings listed and detected a forge --broadcast deploy, and passed 'ls'. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "none: only fixture files, no test suite.",
   "useful": 3,
   "reliable": 3,
   "polish": 3,
   "creative": 4,
   "beyond": 5,
   "total": 23,
   "best": "A pre_tool_call deploy gate that blocks mainnet deploys with open critical/high findings and fails closed, with strict disclosure and no secrets passed to children.",
   "worst": "No tests; a nonexistent path is silently reported as zero findings, and Slither was absent here so Solidity came back as UNCHECKED.",
   "blocked": null
  },
  {
   "pr": 151088,
   "name": "postkitty",
   "validate": "passed (hermes plugins validate: all checks ok, run by tryit.py with Hermes from a fresh clone of NousResearch/hermes-agent)",
   "inside": "Enable and doctor passed; its tools and engine for checking and making social media video/image posts are loaded. Not run through a model here. Planned calls could not run: hermes chat failed here because the sandbox refused the gateway's control socket (Operation not permitted), 0 model requests; judged from doctor, direct calls and reading.",
   "tests": "partial: 223 passed, 25 failed, 5 errors, 6 skipped; the failures I checked were because ffmpeg is not installed on this machine (FileNotFoundError: 'ffmpeg'), after installing its declared python deps.",
   "useful": 4,
   "reliable": 4,
   "polish": 5,
   "creative": 4,
   "beyond": 4,
   "total": 25,
   "best": "A large, polished, well-tested media-checking and making engine (22 test modules, brand docs, CI) that goes well beyond a skill.",
   "worst": "Needs ffmpeg and heavy dependencies (opencv); I read little of its 20k lines of code, so the score leans on tests and README.",
   "blocked": null
  }
 ]
}