17 skills found
rwi001 / llm-evaluationImplement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking
google-gemini / behavioral-evalsGuidance for creating, running, fixing, and promoting behavioral evaluations
refreshdotdev / web-eval-agentAn MCP server that autonomously evaluates web applications.
opendatalab / omnidocbench-eval-helperHelp users deploy, validate, run, and parse OmniDocBench evaluations. Use this skill whenever the user mentions OmniDocBench, document parsing/OCR benchmark scoring, MinerU or other model evaluation on OmniDocBench, CDM formula metrics, end2end/md2md configs, Docker/conda deployment, remote SSH/H-cl…
Duds / accessibilityAn MCP for testing, evaluating and auditing website accessibility.
Evol-ai / SkillCompassEvaluate agent skill quality. Find the weakest link. Fix it. Prove it worked.
PAIR-code / eval-prSet up a git worktree to run and evaluate an upstream Pull Request locally in a flat worktree bare clone repository layout.
greynewell / mcpbrBenchmark your MCP server.
majiayu000 / issue-size-estimationIssue粒度判定とコード行数見積もりの基準、サイズラベル、分割判定ロジックを定義
evalstate / fast-agentCode, Build and Evaluate agents - excellent Model and Skills/MCP/ACP/A2A Support
Huaizz-shawen / check-statusOpenclaw + rosclaw + visual feedback
lucianwhy / final-review用于期末复习、考前突击、题目生成、知识点辅导、复习资料清洗等场景。基于用户上传课程资料开展工作，优先参考历年真题、老师PPT、平时作业和 Markdown 资料；先用 markitdown 转 Markdown，再清洗材料，最后生成题目、答案和偏考试得分导向的解析。
Guilhem-Bonnet / godot-evidence-loopBoucle de preuve Godot via le serveur MCP Fennara : ancrer chaque claim sur des diagnostics éditeur, screenshots, contexte scène et logs runtime réels
qase-tms / qase-mcp-serverAn official Qase MCP server
TimothyVang / verdict-dfirVERDICT — a DFIR agent (Claude Code as the engine) that produces a signed, offline-verifiable verdict. SANS Find Evil! 2026.
synapseradio / question-the-questionExamine whether the inquiry is aimed at the right target
raphaeltm / doEnd-to-end task execution — research, plan, implement, review, and merge via PR
