Project navigator · Browse by task · 完整中文导航
M.S. student in Intelligent Science and Technology at Beijing Institute of Technology (2026 cohort). My research focuses on agent collaboration, agent safety, and multimodal large language model safety. I also build developer tools and applications.
Paper accepted at KSEM (CCF-C) · Huawei ICT Competition: National First Prize
Portfolio · 中文简历 · English Resume · Email
Download: 中文 PDF · English PDF
- KSEM · CCF-C: Paper accepted. CCF classification.
- Research: agent collaboration, agent safety, and multimodal large language model safety.
- Education: B.S. in AI, Guizhou University; rank 1/69, GPA 4.14/5. M.S. at BIT.
- Honors & international experience: National Scholarship; Huawei ICT National First Prize; NUS exchange; IELTS 6.5.
31 original Skills · 33 Python scripts — source inventory checked on 2026-10-30.
- Developed 31 original Agent Skills and 33 Python scripts implementing 23 investment-method reference frameworks across value, growth, momentum, macro cycles and allocation.
- Combined financial screening, factor scoring, cross-framework consensus and cost-aware backtesting, with component scores, turnover, holding details and input hashes.
- The project's validation record documents 60 passing local tests; a one-command offline demo produces reports, CSVs and logs.
- Retrieved real FY2021–2025 financials for 600519 with AKShare 1.19.3 on 2026-10-24.
- Project walkthrough · Run the offline demo · Validation record · Original live-data record
Selected maintenance priorities. Open each first-run guide for inputs, outputs, environment and limits.
| Project | Purpose | Start |
|---|---|---|
| awesome-stock-quant-skills | Connect financial screening, consensus ranking and cost-aware backtest arithmetic. | First run |
| xss-audit-desktop | Keep source locations, bounded flow evidence and human decisions together. | First run |
| agent-trace | Inspect tokens, tool calls and estimated costs from local coding-agent logs. | First run |
| paper-zh-translation | Write supplied Chinese translations into the paper's PDF coordinates and inspect the result. | First run |
| beiligong-ppt-skill | Turn font, background, title and logo changes into a repeatable slide workflow. | First run |
| taotao-futures-research | Compare sectors with explicit common windows, information cutoffs and sample coverage. | First run |
Portfolio / 作品集 · All projects / 全量索引
| Tool | Purpose | Command |
|---|---|---|
| agent-trace | Local agent session analytics | npx @liangzhengtao/agent-trace |
| git-format | Rule-based commit message formatting | npx @liangzhengtao/git-format --dry-run |
| ai-commit | Generate messages from Git diffs | npx @liangzhengtao/commit-ai |
| vibe-check | Check AI coding configuration | npx @liangzhengtao/vibe-check |
AI / Data: Python · PyTorch · Transformers · RAG · Milvus · pandas
Applications: TypeScript · JavaScript · React · Flask · FastAPI · SQLite
Workflows: Git · Node.js CLI · Agent Skills · GitHub Actions
GitHub activity
Interested in research collaboration or AI engineering? Reach me at 1392452425a@gmail.com.