Agent Harness (EvalOps) Model Card
这是可索引项目报告证据页：它保留 Agent Harness (EvalOps) Model Card 的源材料入口、机制线索和限制提醒；正文仍需 reader/editor 与 academic public-copy review 后才能当作最终结论引用。
Agent Harness (EvalOps) Model Card
| Field | Value |
|---|---|
| Repository | evalops/agent-harness |
| Category | Cross-Provider Agent Harness Adapter |
| Stars / forks snapshot | 18 / 5 |
| Commits / issues / PRs snapshot | 12 / 5 / 0 |
| Language | Python |
| License | MIT |
| Raw capture | raw-github/evalops_agent-harness.md |
| Updated by | hourly public metadata update, 2026-06-04 15:36 +0800 |
1. Role in Self Evolve
evalops/agent-harness is a lightweight harness adapter that lets one tool registry and one prompt surface run across OpenAI and Claude agent SDK backends. It matters because self-evolving agents need inspectable memory, skill, harness, benchmark, and runtime substrates before improvement claims become trustworthy.
2. Working Principle
register tools once -> normalize json schema and response shape -> lazy-load provider adapters -> run the same task across multiple agent backends for comparison
3. Evidence Path
web-observed GitHub page showed 18 stars, 5 forks, 12 commits, 5 open issues, 0 pull requests, visible MIT licensing, and a unified harness surface for OpenAI Agents SDK and Anthropic Claude Agent SDK. This iteration keeps freshness honest: the snapshot comes from the public GitHub page observed on 2026-06-04, while shell GitHub API access remained blocked in this workspace.
4. Teaching Use
Use this card to explain Cross-Provider Agent Harness Adapter: it shows how raw public GitHub evidence becomes a project-facing teaching artifact inside the self-evolving-agent pipeline.
5. Limits
The repository was not cloned in this iteration; no benchmark run, workflow execution, or agent loop experiment was executed. Counts and claims are visible public-page signals unless independently revalidated later.