Use or change a skill · Workflow examples
Inventory
Section titled “Inventory”80 skills. One owner per directory; each name is a permalink.
| Name | Owner | Description | Source |
|---|---|---|---|
add-debug-info | installer-npx | Emitting debug info from LLVM (DIBuilder, locations, LLVM 22) — pairs with dwarf-debug-format from the consumer side | packages/agent-skills.txt:155 |
ast-grep | installer-npx | ast-grep: structural search/rewrite — the official skill teaches pattern syntax + rule files; fills the one capability gap rg/sd can’t cover. | packages/agent-skills.txt:22 |
blender-workbench | repository | Create, edit, render, animate, and export Blender scenes using bpy, the live Blender MCP bridge, and visual inspection. Use for Blender authoring, procedural assets, 3D scene work, and game-ready GLB exports. | home/dot_claude/skills/blender-workbench/SKILL.md:1 |
browser-workbench | repository | Build, debug, and visually verify web interfaces with native browser control and repeatable Playwright flows. Use webmcp-workbench when adding or testing page-provided agent actions. | home/dot_claude/skills/browser-workbench/SKILL.md:1 |
compiler-workbench | repository | Design or change language tooling with explicit grammar/IR boundaries, correctness invariants, and phase-by-phase validation. | home/dot_claude/skills/compiler-workbench/SKILL.md:1 |
concurrency-deadlock-remediation | repository | Use when finding and fixing deadlocks with lock ordering, reproduction, timeouts, or lock-free alternatives. Triggers: | home/dot_claude/skills/concurrency-deadlock-remediation/SKILL.md:1 |
cuda | installer-npx | CUDA debug/profile workflow + PTX ISA 9.1 vendored as searchable markdown — grounded PTX lookup instead of hallucination. | packages/agent-skills.txt:114 |
debug | repository | Investigate bugs with an uncertain root cause, conflicting evidence, intermittent reproduction, or failed fix attempts. Use to design discriminating experiments and trace faults; a known, mechanical correction does not require this workflow. | home/dot_claude/skills/debug/SKILL.md:1 |
desktop-workbench | repository | Use when operating or validating native desktop workflows across apps, including macOS background computer use, file dialogs, creative applications, and CLI-driven accessibility automation. | home/dot_claude/skills/desktop-workbench/SKILL.md:1 |
doc-coauthoring | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:190 |
docx | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:48 |
domain-modeling | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:145 |
dwarf-debug-format | installer-npx | DWARF format + dwarfdump/llvm-dwarfdump workflows (user-requested — partly reference-style, kept for the format internals) | packages/agent-skills.txt:151 |
dynamic-analysis | repository | Investigate binary behavior through runtime instrumentation or reverse engineering with QEMU, GDB, Frida, or syscall tracing. Use for emulation, memory inspection, and function interception; ordinary builds, tests, and command execution do not need this skill. | home/dot_claude/skills/dynamic-analysis/SKILL.md:1 |
elf-expert | repository | Provides expertise for inspecting, analyzing, and modifying ELF (Executable and Linkable Format) binaries and understanding the ELF format/ABI. Triggers when examining ELF headers, sections, segments, symbols, relocations, or dynamic data; inspecting binary hardening (NX, RELRO, PIE, canaries); using readelf/objdump/nm; modifying ELF files (strip, objcopy, patchelf); or writing/reviewing code that parses ELF data. | home/dot_claude/skills/elf-expert/SKILL.md:1 |
env-reconciler | repository | Reconcile a repository’s expected environment with the local machine and produce a startup plan. Use when entering an unfamiliar repository, or when build/test commands depend on missing tools, unclear shell state, runtime versions, package managers, secrets, or local services. | home/dot_claude/skills/env-reconciler/SKILL.md:1 |
ffmpeg | installer-npx | ffmpeg: command cheatsheet skill (rendi-api — ffmpeg API vendor) | packages/agent-skills.txt:59 |
find-skills | installer-npx | find-skills (vercel-labs): discover/install registry skills from inside a session — the discovery half of this very file’s workflow. | packages/agent-skills.txt:134 |
frontend-design | repository | Guidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don’t read as templated defaults. | home/dot_claude/skills/frontend-design/SKILL.md:1 |
fuzzing-rust-expert | repository | Use this skill to fuzz open source Rust software projects. | home/dot_claude/skills/fuzzing-rust-expert/SKILL.md:1 |
game-systems | repository | Implement game mechanics and engine systems with clear loop invariants, deterministic update ordering, and rapid gameplay validation. | home/dot_claude/skills/game-systems/SKILL.md:1 |
gdb-for-debugging | repository | Investigate Linux process crashes, hangs, spin loops, deadlocks, thread starvation, and core dumps with GDB, strace, and /proc inspection. Use when stack, thread, memory, or ptrace evidence is needed, not for a routine build or a known test failure. | home/dot_claude/skills/gdb-for-debugging/SKILL.md:1 |
glab | installer-self | Installer-managed agent skill | packages/agent-skills.txt:56 |
google-workspace-flow | repository | Edit shared native Google Docs, Sheets, and Slides with templates, comments, and visual verification. Use for collaborative Google-file authoring and review through an available plugin, Workspace MCP, or gws CLI. | home/dot_claude/skills/google-workspace-flow/SKILL.md:1 |
grill-me | installer-npx | First-principles design interrogation. grill-me/grill-with-docs are small explicit entrypoints; grilling and domain-modeling provide the real workflow. | packages/agent-skills.txt:142 |
grill-with-docs | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:144 |
grilling | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:143 |
history-search | repository | Search past coding-agent sessions (Claude Code, Codex, opencode, pi transcripts) with hybrid lexical+semantic search via cass. Use when the user references past work (“we did this before”, “that bug from last month”), before re-debugging something that feels familiar, or to recover the reasoning behind an old decision. | home/dot_claude/skills/history-search/SKILL.md:1 |
jit-setup | installer-npx | ORC JIT v2 (LLVM 22) + backend-skeleton walkthroughs — ORC churned hard post-training; the model mixes API eras without this. | packages/agent-skills.txt:104 |
jj-guide | installer-npx | jj (Jujutsu): community usage guide. Unpinned: the skills CLI cancels on the lockfile (~/.agents/.skill-lock.json records the tree hash). | packages/agent-skills.txt:30 |
kernel-triton-writing | installer-npx | Triton kernel traps (NVIDIA official; lives in the TensorRT-LLM repo, so the one-time fetch is heavyweight): C-semantics //%, fp32 transcendental casts, autotune decorator order, TF32 defaults. | packages/agent-skills.txt:110 |
lean4 | installer-npx | lean4 shared skill: gives Codex/opencode/pi/Cursor the Lean workflow knowledge. Claude ALSO has the richer lean4@lean4-skills PLUGIN (commands+hooks+agents); the near-duplicate is deliberate — plugin = cycle engine, skill = portable docs. | packages/agent-skills.txt:194 |
llvm | installer-npx | LLVM: new pass manager surface (-passes=‘default | packages/agent-skills.txt:78 |
llvm-development | repository | Conventions for writing, refactoring, or reviewing C++ in LLVM, Clang, or LLDB trees. Covers the naming split between LLVM and LLDB, include ordering, ADT containers and string types (StringRef, SmallVector, DenseMap), Error/Expected handling, LLVM-style casts, formatting rules, and LLDB-specific hazards. Load when editing .cpp/.h files under an LLVM monorepo, opening a file that includes llvm/ADT/, or answering questions about LLVM/LLDB coding standards. | home/dot_claude/skills/llvm-development/SKILL.md:1 |
llvm-howto | installer-npx | LLVM upstream-developer workflow (dtcxzyw, prolific LLVM contributor): alive-tv verification, llvm-reduce, llubi, lit/FileCheck. Complements llvm. | packages/agent-skills.txt:95 |
long-running-work | repository | Carry sustained research and implementation across interruptions, context windows, parallel agents, and remote sessions with resumable project state and verified artifacts. | home/dot_claude/skills/long-running-work/SKILL.md:1 |
macho-expert | repository | Inspect, explain, or modify Mach-O structure, architecture slices, dynamic loading, code signatures, entitlements, and binary hardening; write or review Mach-O parsers. Use for format and loader analysis, not merely running or building a macOS application. | home/dot_claude/skills/macho-expert/SKILL.md:1 |
macprof | repository | Profile macOS processes (CPU hot spots, hangs, “what is X doing?”). Wraps sample/spindump/xctrace and prints a compact text summary or speedscope/flamegraph data. | home/dot_claude/skills/macprof/SKILL.md:1 |
manim-composer | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:41 |
manimce-best-practices | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:42 |
math-lookup | repository | Look up mathematical reference data from authoritative sources — integer sequences (OEIS), L-functions/number fields (LMFDB), reviews (zbMATH Open), DOI→BibTeX, constant identification via PSLQ, and Lean/Mathlib theorem search (LeanSearch/Loogle). Use when identifying a sequence or constant, checking whether a result/object is known, fetching citations, or searching Mathlib from outside a Lean session. | home/dot_claude/skills/math-lookup/SKILL.md:1 |
media-workbench | repository | Create motion graphics, explainers, and editable video projects with Remotion, Kdenlive, and FFmpeg. Use for video authoring, timeline editing, media automation, and rendered delivery. | home/dot_claude/skills/media-workbench/SKILL.md:1 |
metal-gpu-debug | installer-npx | Metal debug/profile with xctrace/.gputrace capture→fix→verify loop — sibling to the vendored cuda skill, for Apple-silicon GPU work. | packages/agent-skills.txt:186 |
mlir | installer-npx | MLIR reference pack: ODS/DRR/PDLL, dialect conversion, Transform dialect — 37 reference files. CAVEAT: pinned to MLIR 19 APIs; verify against current LLVM when the dialect-conversion driver matters. | packages/agent-skills.txt:100 |
mojo-gpu-fundamentals | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:86 |
mojo-syntax | installer-npx | Mojo (Modular official): syntax-churn correction layer + GPU programming. Both are explicit “what pretrained models get wrong” tables — exactly the point of a skill. MLIR-native language, NVIDIA/AMD/Apple-silicon targets. | packages/agent-skills.txt:85 |
motion-designer | installer-npx | Keep the interaction and visual layers distinct from frontend-design. | packages/agent-skills.txt:170 |
native-crash-debug | repository | Apply when debugging a crash, segfault, or bad-access in a compiled binary — C, C++, Objective-C, or any native library. Covers systematic isolation, backtrace-first methodology, dynamic linker diagnosis on macOS (dyld) and Linux (ld.so), and Mach-O / ELF binary inspection. Invoke explicitly with /native-crash-debug or when a segfault, SIGABRT, EXC_BAD_ACCESS, or missing-symbol crash is the presenting symptom. | home/dot_claude/skills/native-crash-debug/SKILL.md:1 |
native-debugger-triage | repository | Use when debugging native programs or ELF binaries with gdb breakpoints, backtraces, and memory inspection. Triggers: | home/dot_claude/skills/native-debugger-triage/SKILL.md:1 |
new-target-backend | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:105 |
nushell | installer-npx | nushell: structured-data pipelines for the installed nu | packages/agent-skills.txt:25 |
pdf | installer-npx | Anthropic’s official document skills (anthropics/skills): read/create/edit PDFs (extract, merge, split, forms, OCR) and Word DOCX (render-and-verify workflow). Google Docs I/O is the google-workspace MCP. | packages/agent-skills.txt:47 |
playwright-cli | installer-npx | Official browser CLI workflow. The CLI itself is already in npm.txt; this gives every harness the same snapshot/ref-driven interaction protocol. | packages/agent-skills.txt:167 |
ponytail-review | installer-npx | Ponytail change review only: concrete findings across correctness, security, load, tests, performance, and unnecessary code; no coding mode or hooks. | packages/agent-skills.txt:138 |
property-based-testing | repository | Provides guidance for property-based testing across multiple languages and smart contracts. Use when writing tests, reviewing code with serialization/validation/parsing patterns, designing features, or when property-based testing would provide stronger coverage than example-based tests. | home/dot_claude/skills/property-based-testing/SKILL.md:1 |
qmd | installer-self | qmd: the tool ships its own skill (targets ~/.agents/skills/qmd, which is the same tree via the symlink) | packages/agent-skills.txt:34 |
research | repository | Research a topic thoroughly using web search before answering. Use when encountering unfamiliar APIs, libraries, frameworks, or domains, when training data may be stale, or when a summary must be anchored in primary sources rather than recall. | home/dot_claude/skills/research/SKILL.md:1 |
router | repository | Choose the project-native tools and installed skills for game development, from gameplay and shaders to assets, editor automation, builds, and playable validation. | home/dot_claude/skills/router/SKILL.md:1 |
rr-debug | repository | Time-travel debugging with rr (Linux) or alternatives (macOS) - record test execution, replay deterministically, step backwards to find root cause of flaky tests and crashes | home/dot_claude/skills/rr-debug/SKILL.md:1 |
rust-skills | installer-npx | Idiomatic-Rust correction layer: 265 rules across ownership, unsafe, async cancel-safety, SIMD/PGO, anti-patterns (Rust 1.96 / 2024 edition) — the Rust analogue of the vercel-react-best-practices correction skills. | packages/agent-skills.txt:121 |
scientific-review | repository | Audit scientific claims, manuscripts, or peer-review records against traceable evidence, or certify completed Lean proofs. Use for literature reviews, review responses, claim-evidence matrices, and final formal certification; use lean4 for routine proof development. | home/dot_claude/skills/scientific-review/SKILL.md:1 |
simulation-lab | repository | Build or modify numerical simulations with explicit state definitions, stability assumptions, and verification against small test cases. | home/dot_claude/skills/simulation-lab/SKILL.md:1 |
skill-creator | installer-npx | skill-creator (Anthropic official): init/validate/package + progressive- disclosure structure guidance — we ship skills from this repo. | packages/agent-skills.txt:127 |
terraform-skill | installer-npx | terraform/opentofu: antonbabenko (terraform-best-practices author) | packages/agent-skills.txt:62 |
tilegym-cutile-python | installer-npx | NVIDIA official: cuTile tile-based kernel programming (TileGym). Siblings (autotuning, triton/julia conversion) exist in nvidia/skills — add on demand. | packages/agent-skills.txt:182 |
tmux | installer-npx | tmux: drive interactive processes in tmux panes (mitsuhiko) | packages/agent-skills.txt:65 |
tree-sitter-grammar | installer-npx | tree-sitter grammar authoring: state-budget engineering, the uint16 action- table ABI limit, precedence ladders — distilled from building a real grammar against 17 reference grammars. Deep content despite the tiny repo. | packages/agent-skills.txt:91 |
tsan | repository | Handle ThreadSanitizer output - redirect to files, summarize races, avoid context flooding | home/dot_claude/skills/tsan/SKILL.md:1 |
typst | installer-npx | typst: typst-cli authoring skill | packages/agent-skills.txt:71 |
uv | installer-npx | uv: Astral’s official skill for uv workflows | packages/agent-skills.txt:68 |
vercel-composition-patterns | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:163 |
vercel-react-best-practices | installer-npx | React/Next.js correction layers distilled from Vercel’s internal guidelines. NB: —skill matches SKILL.md frontmatter name:, which here differs from the repo dir names (react-best-practices/ → vercel-react-best-practices). | packages/agent-skills.txt:162 |
web-design-guidelines | installer-npx | Installer-managed agent skill | packages/agent-skills.txt:171 |
web-shipping | repository | Drive full-stack web work from first principles through a thin vertical slice, with explicit interface contracts and concrete validation commands. | home/dot_claude/skills/web-shipping/SKILL.md:1 |
webgpu | installer-npx | WGSL/WebGPU patterns (thin but real; repo-root skill) | packages/agent-skills.txt:189 |
webmcp-workbench | repository | Build and validate WebMCP site tools for an interactive web application when an agent should call page-native actions instead of guessing DOM controls. | home/dot_claude/skills/webmcp-workbench/SKILL.md:1 |
wolfram-alpha | repository | Queries Wolfram|Alpha for up-to-date computational results and retrieves contextual information via semantic search. Use this skill when the user needs real-world data, calculations, or factual answers about entities in science, math, geography, history, finance, and more. | home/dot_claude/skills/wolfram-alpha/SKILL.md:1 |
wolfram-language | repository | Evaluates Wolfram Language code, searches documentation, inspects code, runs tests, and retrieves symbol definitions. Use this skill when the user needs Wolfram Language computation or development assistance, including symbolic math, data analysis, visualization, or working with .wl/.wls/.wlt files. | home/dot_claude/skills/wolfram-language/SKILL.md:1 |
writing-skills | repository | Improve skill discovery and behavior using realistic tasks in the target harness and model. Use when a skill misroutes, a workflow needs validation, or instruction revisions need comparison. | home/dot_claude/skills/writing-skills/SKILL.md:1 |
zig-0.16 | installer-npx | zig: zigcc community org, current-Zig (0.16) idioms | packages/agent-skills.txt:74 |