poi-agents

Analyzes existing experiment results and designs/executes ablation experiments to verify the contribution of each component.

Skill for writing English text that reads as naturally human-authored, and for reviewing existing English text to detect and fix AI writing tells. Use this skill whenever the user asks to "write naturally", "remove AI tone", "make it sound human", "avoid AI tells", "de-AI", "humanize", "anti-AI writing", "sound less like ChatGPT", or any variation implying they want text free of detectable AI patterns. Also trigger when the user asks to review or edit text for AI-sounding patterns, or when they paste text and ask "does this sound like AI?" Even if the user simply says "write me a blog post" or "write an essay" without explicitly mentioning AI — if this skill is available, apply its principles by default to produce more natural prose.

Brainstorm research ideas and formulate research hypotheses. When a user presents a research topic or area of interest, explore related ideas.

Create a detailed experiment plan for a selected research idea. Define experiment variables, metrics, baselines, and datasets.

Deep literature review with multi-stage search, citation graph traversal, and full-text analysis. Uses Semantic Scholar API, arXiv HTML reading, and snowball search to produce comprehensive research landscape analysis.

Analyzes experiment results and generates publication-quality figures for papers using matplotlib.

Critically reviews a written paper from the perspective of a top-venue reviewer, providing strengths/weaknesses analysis and improvement suggestions.

Executes the full research pipeline sequentially from brainstorming to paper writing. Passes results from each stage to the next.

Writes a systematic rebuttal to reviewer comments. Classifies each comment and determines accept/rebut/partial-accept to generate a persuasive response.

Deploys experiments to remote machines (GPU servers, clusters) via SSH and automates execution/monitoring/result collection. Supports distributed execution across multiple machines by condition.

Write and execute experiment code according to the experiment plan. Read config.yaml to automatically run the experiment and save results.

Interactive setup wizard for new users who cloned this boilerplate project. Checks prerequisites, installs dependencies, configures environment, and introduces the research pipeline.

Writes an academic paper based on experiment results and figures. Generates a publication-ready paper in LaTeX format.