Evergreen Tester — Development agent for Claude Code
Runs one eval case for an evergreen skill in a fresh context and returns a strict trace report (skills invoked, tools called in order with key inputs, files written, the evidence the case asked for, f.
How to install Evergreen Tester
Installs to ~/.claude/agents/m4bwav-evergreen-protocol-evergreen-tester.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/m4bwav/evergreen-protocol/HEAD/agents/evergreen-tester.md -o ~/.claude/agents/m4bwav-evergreen-protocol-evergreen-tester.md Restart Claude Code, or start a new session, for it to be picked up.
What Evergreen Tester does
name: evergreen-tester description: Runs one eval case for an evergreen skill in a fresh context and returns a strict trace report (skills invoked, tools called in order with key inputs, files written, the evidence the case asked for, final message), so the caller can grade on evidence rather than on prose. Use from evergreen-test and evergreen-tune, one case per call; give it the prompt, the run number, the evidence to collect, and whether the skill under test should be treated as absent (b
Alternatives in Development
- Error Trace — You are an error tracking and observability expert specializing in implementing comprehensive error 31.9k ★
- Function Trace Generator Agent 2.3k ★
- Updater — 负责归档 lore-keeping 和规划时设定变更——两种操作流程由 order 类型决定 669 ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Property Auditor
Property-based testing lane (proptest) - lives in the null between hand-written examples by generating valid i
Agentapp Runtime
Agent 应用运行时工程师(Runtime + 可观测性 + API)。当需要实现 agent 编排循环、会话状态管理、流式输出、tool 执行调度、记忆存储,以及最关键的——tracing 可观测性(每一步可回放)与
Log Correlator
Reads logs across many Spring Boot services, correlates them by a trace key (e.g. sessionId), reconstructs one
Optimist
Steel-mans the upside of an idea. Names the specific wins if it works, the compounding effects, the second-ord
Eval Implementer
Implementer role of the D365FO agent eval loop. Runs ON THE VM (mcp-server in full mode + C# bridge) against t
Agentapp Reality Checker
Agent 应用最终验收官。在 eval 循环收敛、安全审查完成后,由 agent-dev-team 剧本在 Phase 7 激活。对照能力画像与 eval 达标情况做整体验收,默认判决是"需要返工",只有压倒性证据才判
Related Skills
Knowledge Ops
Use when explicitly asked to add, edit, delete, list, or sync indexed knowledge sources for Lumina through the
Design Doc
Write a design doc / RFC as Markdown in the user's repository. Use when invoked as /document-design-system:des
Accessibility Specialist
Use when a screen might fail WCAG. Unlabeled inputs, no keyboard path, contrast below AA, missing landmarks, b