Research skills

Browse reusable Agent Skills, each with a clear purpose and practical guidance.

nxv

Find any version of any Nix package across nixpkgs git history using the nxv CLI or HTTP API. Use when asked which nixpkgs commit shipped a specific package version (e.g. "python 2.7", "nodejs 15", "ruby 2.6"), when looking up package metadata/license/homepage, when generating a `nix shell nixpkgs/<commit>#pkg` invocation for an old version, or when querying the public/private nxv server. Triggers include "find python 2.7 in nixpkgs", "which commit had nodejs 15.14", "when was foo added/removed", "give me the nix shell command for ruby 2.6", "search nixpkgs for X", or any question about historical Nix package versions.

133 repo starsObserved in 1 repos
Research
GitHub

paper-cluster-by-theme

Groups a set of kept papers into 2-5 thematic clusters before synthesis, using abstract semantics + matched-keyword overlap. Names each cluster with a short noun phrase that describes what the cluster argues, not just the topic. Surfaces an "outliers" bucket for single-paper themes that don't fit. Domain-neutral - usable for any literature-scan workflow. Use after relevance filtering and before writing the synthesis report. Trigger keywords - cluster papers, group by theme, thematic clusters, paper themes, organize papers.

133 repo starsObserved in 1 repos
Research
GitHub

paper-relevance-filter

Scores a candidate paper against a keyword watchlist and a relevance-criteria document, returning KEEP/DROP/REVIEW with a one-line rationale and a 0-100 relevance score. Combines keyword-match strength, criteria-fit, and a historical-context check (was this paper or its preprint already covered in a recent digest). Domain-neutral - usable for any literature-scan workflow. Use after fetching candidate papers from bioRxiv, medRxiv, or PubMed and before clustering or synthesis. Trigger keywords - paper relevance, filter papers, keep or drop, score papers, relevance rationale.

133 repo starsObserved in 1 repos
Research
GitHub

paper-three-pass-extraction

Domain-neutral methodology for autonomously extracting structured notes from a single academic paper via three escalating passes. Pass 1 (inspectional, ~10-15 min) reads title, abstract, intro, section headings, conclusion, references at a glance, then applies the Five Cs framework (Category, Context, Correctness, Contributions, Clarity). Pass 2 (content grasp, ~30-60 min) reads the full paper skipping proofs, answers main-argument / Big-Question / hypotheses / figure-by-figure / references / confusions. Pass 3 (deep understanding, ~1-4 hours, reserved for important papers) virtually re-implements, challenges every assumption, identifies what is NOT said, asks the falsifiability question. The methodology is internal to the agent applying it - questions are answered against the paper's content, never asked of the operator. Inspired by Keshav 2007 ("How to Read a Paper") and Adler-style inspectional reading. Use when an extraction agent needs to convert dense academic prose into structured machine-and-human-readable notes - bio papers, CS papers, ML papers, statistics, math, any field.

133 repo starsObserved in 1 repos
Research
GitHub

symmetry-discovery-questionnaire

Guides collaborative discovery of hidden symmetries in ML data through structured domain analysis, coordinate system examination, transformation testing, and physical constraint identification. No group theory knowledge required. Use when ML engineers need to identify symmetries in their data, when user mentions data symmetry, invariance discovery, what transformations matter, or needs help recognizing patterns their model should respect.

133 repo starsObserved in 1 repos
Research
GitHub

synthesis-and-analogy

Synthesizes information from multiple sources into coherent insights and applies analogical reasoning to transfer knowledge across domains. Use when conducting literature reviews, integrating stakeholder feedback, reconciling conflicting viewpoints, identifying cross-source patterns, creating explanatory analogies ("X is like Y"), finding creative solutions through cross-domain transfer, or testing whether analogies hold (surface vs deep). Use when user mentions "synthesize", "combine sources", "analogy", "similar to", "transfer from", "integrate findings".

133 repo starsObserved in 1 repos
Research
GitHub

synthesis-application

Domain-neutral methodology for evaluating completeness and logical soundness of an extracted set of components, then transforming them into actionable guidance. Runs the "is it true / is it complete / what of it" critical evaluation pass before any final artifact is built. Checks for completeness gaps, logical consistency, contradictions, and practical applicability. Reusable across any extraction workflow - skill creation (evaluating extracted components before building SKILL.md), paper extraction (evaluating Pass 2 extraction notes before deep reading), report writing (evaluating gathered evidence before synthesis). Use when an agent has extracted structured components from a source and needs to gate-check before downstream commitment. Trigger keywords - synthesis evaluation, completeness check, logic check, critical evaluation, fact-check before synthesis, gap analysis, what is not said.

133 repo starsObserved in 1 repos
Research
GitHub

virtual-memory-paging-and-tlb

Virtual memory skill for paging, page tables, and TLB. Use when explaining page faults, multi-level page tables, TLB behavior, or virtual vs physical addressing. Activates on queries about virtual memory, page table, TLB, page fault, mmap paging, or x86-64 paging.

133 repo starsObserved in 1 repos
Research
GitHub

worked-example-walkthrough

Produces step-by-step computational walkthroughs of vector and matrix operations as a sequence of numbered "frames", showing the explicit state at each step. The text-equivalent of a 3Blue1Brown animation — each frame shows what changed and why, so the learner can re-trace the operation by hand. Use when the learner needs to *see* a computation unfold (eigenvalue computation, attention with 3 tokens, gradient descent step, SVD on a 2×2, layer norm on a 3-vector, softmax of a small input), when an explanation has been given but the learner needs to ground it in a worked example, or when introducing an operation that's intimidating in symbol form but trivial in pencil-and-paper form.

133 repo starsObserved in 1 repos
Research
GitHub

dueling-autoresearch

Use when the user wants two approaches raced head-to-head on a single shared metric — e.g. a classical/algorithmic lane vs an ML/learned lane, or any two strategies for the same task. Each lane runs its own analysis-first research loop confined to its lane, the lanes share a scoreboard and may borrow ideas across the boundary without abandoning their identity, and a shared eval keeps the head-to-head honest; loops until interrupted, reporting the current leader. Not for improving a single approach in isolation (use a single-track research loop), and not for picking between two finished artifacts in one shot (that is a one-time comparison).

132 repo starsObserved in 2 repos
Research
GitHub