gudo7208

Maintain the Awesome Coding Agent Eval data corpus — collect new arXiv papers not yet in the repo, audit and fix data-quality defects (dangling cross-references, duplicate entries, taxonomy drift, title-dump descriptions), and regenerate README/docs/llms.txt. Use for routine repo upkeep, e.g. when the user says "维护仓库 / 更新论文 / 补录论文 / 查数据漏洞 / 数据审计 / weekly update / daily maintenance / refresh the corpus".

Munger's multi-disciplinary mental models for decision-making. Triggers when user faces decisions, analyzes problems, or needs multi-perspective thinking. Uses layered loading: L0 index routing → L2 on-demand model loading. Parallel analysis via Agent Team, each agent handles one mental model perspective, then synthesizes into comprehensive decision support.

Generates project documents in the authentic style of real engineers — not AI-flavored corporate prose. Covers experiment records, progress reports, technical proposals, evaluation analyses, issue analyses, and executive-level technical briefings. Trigger words: project doc, experiment log, progress report, technical proposal, evaluation, write doc, weekly report, boss briefing, technical plan, kickoff

Academic paper deep analysis and research automation. Four modes: 1. Single: Given a paper, perform deep structural analysis 2. Batch: Given multiple papers, parallel processing with cross-paper insights 3. Triage: Quick relevance assessment against research direction, no full notes 4. Topic: Given a research topic, auto-search, filter, analyze each paper, generate survey report Auto-archives to workspace with bidirectional links and index maintenance.

Security audit for Claude Code Skills. Triggers when user asks to review, check, or evaluate a skill's safety. Covers 10-dimension scanning: network behavior, file boundary violations, data exfiltration, prompt injection, credential exposure, dynamic execution, hooks behavior, supply chain dependencies, SSRF/path traversal, persistence backdoors. Quantified scoring per dimension, generates comprehensive security report.