Featured
hermes-agent/hermes
Hermes Agent is a multi-gateway automation agent with a skill and tool system.
assay, n. — the testing of a metal to determine its purity.
How good is your code, really?
Paste a public GitHub repository. Assay reads the evidence — files, features, the snapshot itself — and hands down a scored verdict. Every claim cited. No hand-waving.
Preview deployment: submissions resolve against fixture evaluations from the versioned report contract, not a live analysis engine. The report you see is the real output format.
The verdict you get
Not a star rating. A scored, versioned document with confidence bands, graded evidence, and a hard release gate — a score that cannot be backed by evidence is withheld, never faked as a zero.
specimen · hermes-agent/hermesopenai api engine · rubric project-rubric-1 · revision 01234567
Hermes Agent is a multi-gateway automation agent with a skill and tool system.
Assay Score0/ 100
- Project Substance90
- Originality72
- Engineering Rigor86
- Open Source Readiness84
- Maintenance Health83
- Potential80
7 evidence records back this report — files, repository features, and the snapshot itself, each graded and revision-pinned.
Every score is traceable
Deterministic collectors gather graded evidence from the repository; an AI evaluator judges it against a versioned rubric; a compiler turns the judgments into scores. Walk any number in the report back to the commit-pinned record it stands on. If it can’t be cited, it isn’t scored.
On the right: one real chain from the specimen above — a dimension score, the evidence id it cites, and the graded record behind that id.
Assay evaluates the project, not its authors or their use of AI-assisted development tools. Scores are dimensioned, versioned, and grounded in cited evidence.
- JudgmentEngineering Rigor · 86/100 · project-score-1
- Citesevidence:repository:snapshot
- Evidence recordrepository · grade A · collected 2026-07-14 · revision 0123456789ab
See where you’d land
Anonymous evaluations publish automatically. Featured projects are editorially selected and labeled; featuring never buys a point.
FeaturedProvisional
openclaw/openclaw
OpenClaw is an autonomous agent runtime with a command-line entry point.
A score-range filter lists only released scores; unavailable and insufficient results are not treated as a zero.
Recently Assayed
- hermes-agent/hermes
Hermes Agent is a multi-gateway automation agent with a skill and tool system.
cli_developer_toolagentautomationdeveloper_tool - openclaw/openclaw
OpenClaw is an autonomous agent runtime with a command-line entry point.
applicationagentgame - example-org/sample-project
The project ships a command-line interface with JSON output.
cli_developer_tooldeveloper_tool - acme/recently-analyzed
The project ships a command-line interface with JSON output.
cli_developer_tooldeveloper_tool
Leaderboard
Ranked by released Assay Score, with evaluation version, engine, confidence, and provisional state in plain sight. Unreleased scores do not rank — nobody gets a free zero.
- hermes-agent/hermes
Hermes Agent is a multi-gateway automation agent with a skill and tool system.
cli_developer_toolagentautomationdeveloper_tool - acme/recently-analyzed
The project ships a command-line interface with JSON output.
cli_developer_tooldeveloper_tool - example-org/sample-project
The project ships a command-line interface with JSON output.
cli_developer_tooldeveloper_tool - openclaw/openclaw
OpenClaw is an autonomous agent runtime with a command-line entry point.
applicationagentgame