Agent skills for research work — autonomous iteration, scientific visualization, bibliography management, image generation, and more. Currently packaged as a Claude Code plugin, but the skills themselves are just markdown files that any coding agent can use.
Read each skill before you run it. These skills shape how your agent works — some, like ralph-loops, run agents with full permissions in autonomous loops. Understand what a skill does before activating it.
These reflect one person's workflow, not universal best practices. Fork the repo, delete what doesn't fit, rewrite what almost fits. The best skills are the ones you've made your own.
# Claude Code
claude plugin marketplace add cailmdaley/skills
claude plugin install research-skillsFresh-context loops: survey, contribute, exit, repeat. Each iteration gets the spec as system prompt and starts clean. Give it a markdown spec describing the desired state, and it loops in a tmux session until reality matches — attach to watch, steer, or update the spec mid-run.
Tell Claude to "use the ralph-loops skill to help me write a spec for [your goal]." Pattern adapted from Geoffrey Huntley's Ralph Wiggum.
A decision framework synthesizing Bertin, Cleveland, Tufte, Cairo, Wilke, and Knaflic. Routes from data type to visualization form, with cosmology-specific conventions (power spectra, covariance matrices, triangle plots, sky maps). Includes a visualization catalog, color palette reference, and design system.
Opinionated approach to research code: conceptual density, concise conditionals, natural line breaking. Write it right the first time — clean, correct, no post-hoc cleanup needed.
Read full paper source from arXiv, fetch BibTeX from NASA ADS, generate citation keys, append to your .bib file. Requires $ADS_API_TOKEN.
Generate and edit images using Gemini CLI with persistent visual memory via KV cache warmth. Two warm caches (Gemini's KV cache + Claude's context) enable iterative refinement. Includes transparency extraction via difference matting.
Requires Gemini CLI with the nanobanana extension.
Speak text aloud using pocket-tts server with streaming audio. macOS only.
Requires: pip install pocket-tts, brew install ffmpeg, a voice file at ~/.config/pocket-tts/default-voice.wav.
MIT