Skip to content

agent.md

Latest commit

 

History

History
52 lines (43 loc) · 2.2 KB

agent.md

File metadata and controls

52 lines (43 loc) · 2.2 KB

Agent Profile

Purpose

Assist with quantitative research notebooks, statistical validation, and clear academic-style writing for ML/quant finance.

Environment

  • OS: Windows
  • Shell: PowerShell
  • Repo root: C:\Users\adamd\workspace\quant_research
  • Active file focus: dont_get_fooled_by_chance.ipynb
  • Notebook format: Jupyter .ipynb

Tools and Workflow

  • Use ripgrep (rg) for fast search; fall back only if unavailable.
  • Prefer Python scripts for structured edits in notebooks.
  • Keep edits minimal and targeted; avoid touching unrelated cells.
  • Validate code paths in-place; do not add auto-formatters unless asked.

Style and Writing

  • Academic, precise, and direct; avoid fluff.
  • Favor terms: ?candidate model?, ?in-sample evaluation?, ?sampling variability?.
  • Frame results in ML pipeline terms when relevant (AutoML, hyperparameter search).
  • Explicitly note limitations (IID assumptions, non-stationarity, economic significance).
  • Keep math notation consistent; keep frequency/annualization explicit.

Technical Preferences

  • Use ASCII by default; only introduce non-ASCII if file already uses it.
  • Add brief comments only where logic is non-obvious.
  • In notebooks, keep markdown concise and structured.

Testing and Validation

  • Run notebook cells only when requested or when validating critical changes.
  • For statistical routines, add simple sanity checks if requested.
  • Avoid long-running simulations unless requested.

Dependency Management

  • Prefer existing requirements.txt or environment.yml.
  • Installing packages should be explicit and minimal; document any new dependencies.
  • Use pip only when asked or required to reproduce a bug.

File and Output Conventions

  • Keep outputs deterministic (seed RNG where applicable).
  • Use clear section headers and descriptive cell titles.
  • For tables/plots, keep labels explicit about units and sampling frequency.

Linting (Ruff)

  • Use Ruff for Python linting when requested.
  • Prefer ruff check . for repo-wide checks.
  • For auto-fixes, use ruff check . --fix only when explicitly asked.
  • Install via pip install ruff (or add to requirements.txt) when setup is requested.

Specialized Agents

  • See AGENTS.md for task-specific agents.