Claude Skills
The open format is called Agent Skills and works in Claude Code, Codex, Cursor and other agents — most people know it as Claude Skills.
Every Agent Skill we could find on GitHub, deduplicated by content. 86 961 files from 2 125 authors, of which 68 678 are unique — the rest is the same skill repackaged into someone else's repository. For each one: what it weighs in tokens, whether it ships runnable scripts, and which MCP servers it needs.
Collect what is new in AI in the last 24 hours across X, Hugging Face, GitHub and sitemaps, deduplicated, as one report of names that are starting to surge. Use when the user asks what is new or trending in AI today, wants a daily AI listening report, or needs fresh names of models, products and repositories to feed keyword research. Runs the new-term-radar command line; needs an AISA_API_KEY.
Use the InfiniSynapse CLI (`agent_infini`) for multi-turn AI data-analysis tasks, database/RAG context, and task workspace files. Use when the user mentions InfiniSynapse, agent_infini, database or RAG analysis, or asks to delegate analysis through InfiniSynapse.
> Predict protein structure for monomers and multimers with AlphaFold2 via the ColabFold runner (Mirdita et al. 2022, github.com/sokrypton/ColabFold; AlphaFold2 Jumper et al. 2021). Reach for this skill to fold a sequence or complex with the AF2/AF2-Multimer evoformer, to validate designed sequences by self-consistency pLDDT, ipTM, and RMSD, or to run a quick MSA-backed prediction using the public MMseqs2 server.
Organize multi-step scientific analyses into reproducible, self-contained modules. Use for workflows such as QC→PCA→DEG→GSEA that produce scripts, inputs, figures, tables, and methods. Creates a stable module layout, records exact inputs/parameters/package and database versions in each module README, keeps large data as references instead of copies, and verifies outputs before completion.
> 给一句话或一段话，找出真实学术文献来反对它——相反结论、边界条件、替代解释、方法批评、复制失败。按威胁程度排序，每篇附一行"它如何威胁这个观点"和"怎么回应它"。底层走真实检索（scimaster-cli），绝不编造反例。 **以下情况请主动触发本技能**：用户想"提前挡住审稿人"、"找这个结论的反例"、"这个方向有没有争议"、"有没有跟这个相反的研究"、"帮我攻击这个论点"、"这个结论稳不稳"、"想知道有哪些反对意见"——即使用户没有说"bear-counter"，只要意图是**为一个观点寻找学术反对证据**，就使用本技能。bear-support 和 bear-counter 配对使用效果最好：对同一段话各跑一遍，就能同时看到正反两面。 不适用于：找支持文献（用 bear-support）、选题查重（用 bear-scoop）、概念地图（用 bear-map）、溯源演化史（用 bear-trace）。
> 给一个概念，从真实检索到的论文摘要里挖出它的邻近概念，画出以该概念为中心的知识地图——每个节点都锚定真实文献，不从记忆里补节点。同时输出 Mermaid 概念图（可在 Claude Code / GitHub / Obsidian 渲染）和可截图的独立 HTML 地图，再给 3–6 篇入门推荐。 **以下情况请主动触发本技能**：用户说"帮我画一张这个概念的知识地图"、"这个词周围有哪些相关概念"、"梳理一下这个方向的概念网络"、"给我一张概念图"、"这个领域的核心概念有哪些"、"知识地图"、"概念地图"、"map out this concept"、"draw a knowledge map of X"、"what concepts surround X"、"show me the concept network around this term"、"concept map for X"——即使用户没有说"bear-map"，只要意图是**以一个概念为中心，了解它周围的知识版图**，就使用本技能。 不适用于：为观点找支持文献（用 bear-support）、找反对文献（用 bear-counter）、选题查重（用 bear-scoop）、溯源演化史（用 bear-trace）。
> 给一个概念或领域，同时画出它的概念版图（空间：现在这个领域长什么样）和演化脉络（时间：这个领域是怎么走到今天的），在一份综合报告里帮你快速建立对一个陌生领域的立体认知。底层走真实检索（scimaster-cli），绝不编造引用。 **以下情况请主动触发本技能**：用户想"快速入门一个领域"、"搞懂这个方向"、"这个领域的核心概念和发展历史"、"帮我建立对这个方向的认知框架"——即使没有说"bear-onboard"，只要意图是**对一个陌生领域同时理解概念版图和演化脉络**，就使用本技能。 不适用于：只画概念地图（用 bear-map）、只做溯源（用 bear-trace）、为观点找文献（用 bear-support / bear-counter）、选题查重（用 bear-scoop）。
> 给一个研究 idea，依次执行撞车检测、安静区支撑检索、潜在挑战检索，在一份综合报告里呈现这个 idea 的立项前证据格局——撞车风险在哪、安静区里有没有文献支撑、最需要提前想清楚的挑战是什么。底层走真实检索（scimaster-cli），绝不编造引用。 **以下情况请主动触发本技能**：用户想"开题前评估一下"、"这个 idea 值不值得做"、"立项前做个全面的文献调研"、"帮我评估这个研究方向"——即使没有说"bear-propose"，只要意图是**在正式投入一个研究方向之前做综合评估**，就使用本技能。 不适用于：只查撞车（用 bear-scoop）、只找支持（用 bear-support）、只找反对（用 bear-counter）、概念地图（用 bear-map）、溯源演化史（用 bear-trace）。
> 给一个观点，同时检索支持它的文献和反对它的文献，在一份综合报告里呈现正反两侧的证据格局——支持侧最强证据是什么、反对侧最危险的挑战是什么、哪些角度有支持但无反对、哪些角度有反对但无支持。底层走真实检索（scimaster-cli），绝不编造引用。 **以下情况请主动触发本技能**：用户想同时看到"支持和反对"、"这个观点稳不稳"、"帮我做一个全面的论点评估"、"正反两面都找一找"——即使没有说"bear-review"，只要意图是**对一个观点做完整的证据格局分析**，就使用本技能。 不适用于：只找支持（用 bear-support）、只找反对（用 bear-counter）、选题查重（用 bear-scoop）、概念地图（用 bear-map）、溯源演化史（用 bear-trace）。
> 在你投入一个研究方向之前，先看看有谁已经在做同样的事。把 idea 拆成多个不同角度分别去检索，返回一张"邻近程度地图"，重点标出最危险的撞车候选。底层走真实检索（scimaster-cli），不做主观新颖性判断。 **以下情况请主动触发本技能**：用户说"这个选题有没有人做过"、"我会不会被抢发"、"这个方向还有空间吗"、"我的 idea 新不新"、"要不要查一下有没有撞车"、"这个想法是不是已经有人发了"——即使用户没有说"bear-scoop"，只要意图是**在提交或动笔前确认 idea 的独特性**，就使用本技能。加 `-r` flag 只看近两年和预印本，适合担心被抢发的场景。 不适用于：为已有观点找支持文献（用 bear-support）、找反对文献（用 bear-counter）、概念地图（用 bear-map）、溯源演化史（用 bear-trace）。
> 给一句话或一段话，找出真实学术文献来支持它，每篇文献附一行"它如何支撑这个观点"的说明。底层走真实检索（scimaster-cli），绝不从记忆编造引用——搜不到就如实说搜不到。 **以下情况请主动触发本技能**：用户想让某个观点"更有依据"、"更经得起质疑"、"加上文献支撑"、"找几篇论文"、"这个说法有没有研究"、"帮我查一下这个结论"、"审稿人会不会说没有引用"——即使用户没有说"bear-support"或"找文献"这几个字，只要意图是**为一个观点寻找学术支撑**，就使用本技能。 不适用于：找反对文献（用 bear-counter）、选题查重（用 bear-scoop）、概念地图（用 bear-map）、溯源演化史（用 bear-trace）。
> 给一个主题或一篇种子论文，向前挖掘它建立在哪些前驱工作上（逐层回溯），再向后补充最新进展，讲清楚这个问题从源头到现在是怎么演化的。底层走真实检索（scimaster-cli），每个节点都锚定真实检索到的论文。 **以下情况请主动触发本技能**：用户问"这个问题是怎么来的"、"帮我溯源这个方向"、"这个领域怎么发展的"、"这篇论文是建立在谁的基础上的"、"从头讲一下这个方向的历史"、"入门一个新领域想先了解演化脉络"——即使用户没有说"bear-trace"，只要意图是**理解一个研究问题或领域的来龙去脉**，就使用本技能。加 `-d N` 控制向前溯源的层数（默认 3 层）。 不适用于：为观点找支持文献（用 bear-support）、找反对文献（用 bear-counter）、选题查重（用 bear-scoop）、概念地图（用 bear-map）。
> Structure prediction for protein, nucleic-acid, and small-molecule complexes with Boltz-2 (Passaro & Wohlwend et al. 2025, github.com/jwohlwend/boltz). Reach for this skill to validate designed binders against a target, to co-fold a protein with a SMILES or CCD ligand, or to get an open-source AlphaFold3 alternative with optional binding-affinity prediction.
> Predict genome-wide functional tracks (RNA-seq, CAGE, DNase, ChIP) from DNA (1) Scoring the regulatory effect of a variant on expression/accessibility, (2) Generating predicted coverage tracks for a locus, (3) Prioritising non-coding variants by predicted track delta.
Use this skill to drive the user's real, persistent Chrome/Chromium session — open pages, read them, click, fill and submit forms, navigate, switch tabs, or scrape content that needs the user's existing cookies and login state. Triggers when the user asks to do something in their browser, log into a site and act inside it, fill out a web form, click through a flow, or extract data from a page that requires being signed in. Tools: browser_setup (check/connect the extension), web_open_tab (open a URL), web_scan (read visible content + actionable elements with ready-made selectors), web_execute_js (click/type/navigate, or a JSON command for tabs/CDP), web_screenshot (see what the tab is showing — layout, charts, canvas, QR codes). Not for the built-in read-only web fetch — this is for interacting with a live browser.
> Structure prediction for protein, nucleic-acid, and small-molecule complexes with the Chai-1 foundation model (Chai Discovery 2024, github.com/chaidiscovery/chai-lab). Reach for this skill to predict an antibody-antigen or protein-ligand complex from a single FASTA, to re-fold designed binders as an AlphaFold-multimer alternative, or to drive co-folding from Python for batched campaigns on a GPU.
Set up and validate a reproducible Python or R environment on a Wisp execution context. Use for a selected local, WSL, or direct SSH context when installing scientific packages, configuring caches, recording interpreter activation, or producing an environment smoke test. Do not use for scheduler clusters or managed cloud providers that Wisp cannot track yet.
Create a Wisp specialist or author a project-local skill using the tools Wisp actually exposes. Use when the user wants a custom Agent persona, a restricted specialist loadout, a new skill, or changes to an existing project-local skill.
> Predict small-molecule binding poses with DiffDock-L (Corso et al. 2023/2024, github.com/gcorso/DiffDock) — blind diffusion docking that places a ligand into a protein pocket without a predefined search box and ranks the samples with a learned confidence model. Reach for this skill to dock a SMILES or SDF against a PDB, to generate ranked 3D poses for a small fragment library, or to get a starting pose for downstream rescoring. DiffDock predicts geometry, not affinity.
> Biohub ESMFold2 / ESMFold2-Fast all-atom co-folding (Candido et al. 2026, github.com/Biohub/esm). Single-sequence and MSA modes; protein, DNA, RNA, ligand (CCD/SMILES), modified residues. FoldBench Ab-Ag 50-55%, PPI 70-77% DockQ-pass. Also covers the ESMC-{300M,600M,6B} protein language models from prediction, and the SAE interpretability head. MIT-licensed weights on structures with single-sequence input, (2) Validating designed binders with ESMFold2-Fast, (3) Running ESMFold2 with MSA input, (4) Getting ESMC embeddings or per-residue mutation scores, (5) Choosing kernel backend and sampling-step settings for paper-faithful throughput.
> Score, embed, and generate DNA sequences with Evo 2, a long-context genomic (1) Computing per-nucleotide or per-sequence likelihoods for variant effect scoring, (2) Embedding genomic windows for downstream classification, (3) Generating DNA conditioned on a prefix, (4) Scoring regulatory or coding regions across species.
> Embed proteins with Meta AI's ESM-2 (`fair-esm` package). Use this skill ML, (2) Masked-LM likelihood / mutation effect scoring, (3) Contact prediction from a sequence.
Compose or improve a publication-grade multi-panel scientific figure from a claim, concrete data paths, or an existing image. Use for figure outlining, parallel panel rendering, exact-grid composition, visual inspection, and adversarial figure review. Use figure-style for one standalone plot and paper-narrative for whole-paper figure ordering.
Publication-grade figure correctness and legibility rules. Load before drawing any plot and call `apply_figure_style()` — sets a role-mapped font-size ladder, outward ticks, frameless legends, and 300-dpi output. The skill is a checklist, not a house look: data fidelity (claim-titles tested against every row, excluded data never enters summaries), label economy (floor and ceiling), colour threading, chart-choice-by-data-shape, layout, and a render-then-verify QA loop (bbox collision + per-panel perceptual check). Ships helpers: focal_palette, bar_with_points, strip_with_median, end_of_line_labels, panel_letter, set_frame, panel_crops. For multi-panel figures load `figure-composer`; for whole-paper figure arc load `paper-narrative`.
> Generate a therapeutic indication dossier. Covers the patient population, epidemiology, disease biology, standard of care, regulatory precedent, and landmark clinical trials.
Use this skill whenever the user provides a scientific paper PDF and asks for a group-meeting literature report, journal-club slides, 文献汇报PPT, 组会PPT, paper presentation, article walkthrough, or to explain a paper with PowerPoint. The skill first reconstructs the paper's scientific logic, then builds author/background sections, chooses an evidence-driven slide outline, crops only main-text figure panels from the PDF, and creates an academic PPT with 10–30 slides including title and conclusion/discussion. Always use this skill for '把这篇论文做成PPT', '文献汇报', 'journal club', or '组会汇报', even if the user only says they need slides.
> Inverse-fold a backbone with ligand, nucleic-acid, and metal context using LigandMPNN (Dauparas et al. 2023, github.com/dauparas/LigandMPNN). Reach for this skill to redesign the residues lining a binding pocket around a bound small molecule or cofactor, to design metal-coordinating sites where the geometry must be respected, or to get threaded designed-sequence PDBs out of any MPNN run.
Find, verify, and synthesize scientific literature — from "what's the seminal paper for X" through full multi-source reviews. Covers grounding claims in real retrieved sources, avoiding fabricated citations, handling retractions, and calibrating confidence to evidence strength.
Configure the local wisp-science runtime — uv/Python bootstrap, Node+scimaster-cli for bear-* literature skills, pixi for bioinformatics multi-env analysis. Detect mainland-China network and apply mirrors. Use when Capabilities shows missing Python/uv/Node/sci/pixi, bootstrap errors, or the user asks to 配置环境 / install Python / uv / Node / pixi / set up the local environment. Not for remote GPU/SSH compute (use compute-env-setup).
Explain Wisp's current managed-model endpoint boundary and plan a safe integration. Use when the user asks to register, start, stop, tunnel, authenticate, or manage a persistent inference service.
> Structure prediction using OpenFold3, an open-weights PyTorch reproduction of AlphaFold3 from the AlQuraishi Lab. Use this skill when predicting protein/nucleic-acid/ligand complex structures with an Apache-2.0-licensed AF3 reimplementation.
Judge and reshape the story told by a manuscript and its figure deck. Use when revising paper structure, testing whether Figure 1 is a hook, ordering figures, moving panels, identifying missing analyses, or defining the claim passed to figure-composer.
Use this skill when the user has attached a PDF, paper, report, or other document and the answer needs its content: summarize a section, compare sections, read specific pages, check the table of contents, or read a value off a figure. The `read` tool cannot parse PDF binary — python is the extraction path. Provides `pdf_pages` (pages as text or rendered PNGs, cached) and `pdf_outline` (embedded-bookmark TOC) in the persistent python kernel; load them once via the Kernel Sidecar exec line that `use_skill` appends. For PDF creation/manipulation, use reportlab/pypdf directly.
Inspect a registered execution server before compute planning and interpret its persisted capability profile. Use when a server is added, when the user clicks Probe, before enabling an unfamiliar SSH/WSL resource, or when deciding whether GPU, sudo/root, a scheduler, Python, R, conda, mamba, or environment modules are available.
> Inverse-fold a protein backbone (PDB structure) into amino-acid sequence with ProteinMPNN (Dauparas et al. 2022, github.com/dauparas/ProteinMPNN). Reach for this skill to run sequence design on RFdiffusion backbones, to redesign one chain of a PDB while holding interface residues fixed, or to generate a temperature-swept set of sequences for downstream folding.
Explain Wisp's current Modal boundary and migrate a requested Modal workload to a supported direct SSH Run when possible. Use when an existing workflow mentions Modal, BYOC cloud compute, provider images, or cloud GPU dispatch.
Submit recoverable SSH-direct research Runs with live progress cards and model-free monitoring.
> Embed and annotate single-cell expression data with scGPT, a foundation model (1) Producing cell embeddings from an AnnData for clustering/integration, (2) Zero-shot or fine-tuned cell-type annotation, (3) Gene-level representation for perturbation/GRN tasks. For probabilistic single-cell models (scVI etc.), use the scvi-tools library.
> Probabilistic single-cell RNA-seq with scvi-tools — scVI for a batch-corrected latent space, scANVI for semi-supervised label transfer, and Bayesian differential expression. Reach for this skill to integrate scRNA-seq batches, embed cells for clustering, transfer annotations from a reference onto a query, or score differentially expressed genes per cluster. For spatial deconvolution / mapping use the cell2location, DestVI, or Tangram methods instead.
Wisp-science's actual agent tool surface and runtime boundaries. Load this when deciding which Wisp tool can perform a task, checking whether Python can reach agent or desktop capabilities, choosing between interactive analysis and persisted Runs, or answering questions about delegation, images, skills, memory, artifacts, lineage, credentials, session history, and other self-introspection capabilities.
Create, update, validate, and evaluate Wisp skills. Use when authoring a project-local or installable skill, refining its trigger description, adding deterministic scripts or Python sidecars, or testing whether another Agent can follow the workflow.
> Inverse-fold a backbone with SolubleMPNN — ProteinMPNN retrained on a soluble-PDB subset (Dauparas et al. 2022) — for sequences biased toward cytosolic expression and reduced aggregation. Reach for this skill when designs from vanilla ProteinMPNN are aggregating or going to inclusion bodies, when redesigning a membrane-adjacent fold for soluble expression, or when an E. coli expression screen is the next step.
Invoke an already configured model endpoint from a supported Wisp execution context and capture the bounded inference as a Run. Use only when the endpoint URL and authentication are already available inside that context; this skill does not register or manage services.
Extract text from Microsoft Word documents
Extract text from images using a vision LLM
Ingest Markdown and plain text files
Extract text from PDF documents
Extract text from Microsoft PowerPoint presentations
Extract conversation turns from AI session history files (.jsonl)
Fetch and extract text from web URLs
Search the web and ingest results as wiki pages
Extract data from Excel and CSV files
Extract transcripts from YouTube videos via the YouTube caption system
Expert patterns for 2D animation in Godot using AnimatedSprite2D and skeletal cutout rigs. Use when implementing sprite frame animations, procedural animation (squash/stretch), cutout bone hierarchies, or frame-perfect timing systems. Trigger keywords: AnimatedSprite2D, SpriteFrames, animation_finished, animation_looped, frame_changed, frame_progress, set_frame_and_progress, cutout animation, skeletal 2D, Bone2D, procedural animation, animation state machine, advance(0).
Expert patterns for 3D level design using GridMap with MeshLibrary, CSG constructive solid geometry, occlusion, and runtime GridMap builders. Use when building 3D levels, modular tilesets, or BSP-style geometry. For sky/fog/Environment recipes, route to godot-3d-lighting. Trigger keywords: GridMap, MeshLibrary, set_cell_item, get_cell_item, map_to_local, local_to_map, CSGCombiner3D, CSGBox3D, CSGSphere3D, CSGPolygon3D, OccluderInstance3D, bake CSG.
Expert patterns for Godot 3D PBR materials using StandardMaterial3D including albedo, metallic/roughness workflows, normal maps, ORM texture packing, transparency modes, and shader conversion. Use when creating realistic 3D surfaces, PBR workflows, or material optimization. Trigger keywords: StandardMaterial3D, BaseMaterial3D, albedo_texture, metallic, metallic_texture, roughness, roughness_texture, normal_texture, normal_enabled, orm_texture, transparency, alpha_scissor, alpha_hash, cull_mode, ShaderMaterial, shader parameters.
Expert patterns for RPG/action ability systems including cooldown strategies, combo systems, ability chaining, skill trees with prerequisites, upgrade paths, and resource management. Use when implementing unlockable abilities, character progression, or complex skill systems. Trigger keywords: PlayerAbility, AbilityManager, cooldown, SkillTree, SkillNode, prerequisites, can_use, execute, ComboSystem, ability_chain, global_cooldown, charge_system, upgrade_path.
Expert patterns for Godot 2D physics including collision layers/masks, Area2D triggers, raycasting, and PhysicsDirectSpaceState2D queries. Use when implementing collision detection, trigger zones, line-of-sight systems, or manual physics queries. Trigger keywords: CollisionShape2D, CollisionPolygon2D, collision_layer, collision_mask, set_collision_layer_value, set_collision_mask_value, Area2D, body_entered, body_exited, RayCast2D, force_raycast_update, PhysicsPointQueryParameters2D, PhysicsShapeQueryParameters2D, direct_space_state, move_and_collide, move_and_slide.
Expert patterns for simplifying 3D games to 2D including dimension reduction strategies, 2.5D fake-depth, isometric ports, camera flattening, physics conversion, 3D-to-sprite art pipeline, and control simplification. Use when porting 3D to 2D, building 2.5D / isometric / fake-depth gameplay, creating 2D versions for mobile, or prototyping. Trigger keywords: CharacterBody3D to CharacterBody2D, Camera3D to Camera2D, Vector3 to Vector2, flatten Z-axis, 2.5D, isometric, fake depth, Y-sort, simulated Z, orthogonal projection, 3D to sprite conversion, performance optimization.
Expert patterns for Godot 3D lighting including DirectionalLight3D shadow cascades, OmniLight3D attenuation, SpotLight3D projectors, VoxelGI vs SDFGI, and LightmapGI baking. Use when implementing realistic 3D lighting, shadow optimization, global illumination, or light probes. Trigger keywords: DirectionalLight3D, OmniLight3D, SpotLight3D, shadow_enabled, directional_shadow_mode, directional_shadow_split, omni_range, omni_attenuation, spot_range, spot_angle, VoxelGI, SDFGI, LightmapGI, ReflectionProbe, Environment, WorldEnvironment.
Claude Skills — questions
Answers built from the skills we actually parsed.