{"batch":"deep-57","setup":"Hermes was cloned from NousResearch/hermes-agent and installed with uv sync --locked using uv-managed CPython 3.14. The prescribed tryit.py harness was attempted, but its fresh-home subprocess stalled before reports were produced; catalog validation was run directly for every plugin. No credentials, model or account were used.","plugins":[{"pr":153365,"name":"lumina","validate":"passed: manifest, loadability, capability probe, declared hooks/tools, security scan and no-core-override checks all passed","inside":"loaded by the Hermes capability probe; registers 10 photo/privacy tools plus hooks and a skill. The scripted tryit chat/calls did not complete because the harness stalled before producing a report. macOS Photos/Vision dependencies and hardware were unavailable here.","tests":"not run to completion: collection hit missing fastapi in test_plugin_api.py; no reliable count","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"rank":2,"best":"A polished on-device Photos privacy product that adds real Hermes tools and leak-prevention hooks, with unusually strong documentation and fixtures.","worst":"It is macOS-only and the runtime path could not be exercised here, while the test suite could not collect without extra dependencies.","blocked":null},{"pr":151411,"name":"salts-secrets","validate":"passed, with warnings: manifest declared public_notes_lookup and several hooks not observed by the probe; security scan reported caution for ssh_dir_access","inside":"loaded by the Hermes capability probe; registers gateway/profile isolation, dashboard/CLI integration, a public-context tool, guest routing and lifecycle hooks. The scripted tryit chat/calls did not complete because the harness stalled before producing a report.","tests":"not run to completion: 14 collection errors because the standalone checkout lacked Hermes gateway imports; no pass/fail count","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"rank":3,"best":"It changes Hermes's gateway behavior so strangers receive an isolated public profile rather than the owner's private context, with approvals and audit surfaces.","worst":"The validator found declared-versus-registered capability mismatches and a security-scan caution, and the integration tests require Hermes gateway modules.","blocked":null},{"pr":152408,"name":"rich-paste","validate":"passed, with warnings for private core imports in benchmark-only code","inside":"loaded by the Hermes capability probe; registers llm_request middleware and a Desktop surface for HTML/Markdown conversion and interleaved image attachments. The scripted tryit chat/calls did not complete because the harness stalled before producing a report.","tests":"failed/not completed: pytest output showed substantial progress with many failures and skips, but the interrupted run produced no final count","useful":4,"reliable":4,"polish":5,"creative":4,"beyond":5,"total":27,"rank":5,"best":"A finished Desktop workflow solves a concrete Hermes weakness by preserving document structure and image position when pasting rich content.","worst":"The validator flags private-core imports and the available test run showed many failures before it was interrupted.","blocked":null},{"pr":153766,"name":"hermes-ussd","validate":"failed: capability probe; register() raised No module named 'aiohttp' (manifest/loadability/security checks passed)","inside":"not loaded in the catalog probe because aiohttp was unavailable; it declares a platform adapter, CLI commands and SMS/USSD skill. The scripted tryit chat/calls were not run after the failed load probe; no Africa's Talking credentials were available.","tests":"not run: runtime dependency/account-backed platform was unavailable","useful":4,"reliable":2,"polish":4,"creative":5,"beyond":5,"total":25,"rank":6,"best":"It brings Hermes to feature phones through a real USSD-to-SMS platform, with allowlisting, an outbox and rate limits.","worst":"The catalog admission probe cannot load the plugin without its declared aiohttp dependency, so the core platform path was not verified.","blocked":null},{"pr":150873,"name":"skill-court","validate":"passed, with warnings for private core imports in appeal.py","inside":"loaded by the Hermes capability probe; registers skill-lifecycle/tool/LLM hooks, middleware, auxiliary work and a Desktop/dashboard surface. The scripted tryit chat/calls did not complete because the harness stalled before producing a report.","tests":"not run to completion: the batch test process was interrupted by the environment/network wrapper before a final count","useful":4,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":28,"rank":4,"best":"It gives Hermes a novel feedback loop that experimentally measures skills, quarantines harmful ones and supports evidence-based appeals.","worst":"Its most important automation is background/gateway-coupled and uses private Hermes imports, so upgrade resilience is less certain.","blocked":null},{"pr":150293,"name":"procgen-studio","validate":"passed: manifest, capability probe, provider registrations, dependency declaration, security scan and desktop surface checks passed","inside":"loaded by the Hermes capability probe; registers image/video generation providers, commands, a skill and Desktop surface, with no model-facing tool declaration. The scripted tryit chat/calls did not complete because the harness stalled before producing a report.","tests":"not run to completion: no final count was produced by the interrupted environment run","useful":5,"reliable":4,"polish":5,"creative":5,"beyond":5,"total":29,"rank":1,"best":"It changes Hermes's existing image-generation capability into a local, editable, reproducible SVG/Python rendering pipeline with a critic loop and Desktop integration.","worst":"The provider path is comparatively complex and depends on rendering/tooling such as Cairo, numpy, Pillow and optionally ffmpeg, none of which was exercised end to end.","blocked":null}]}