A curated collection of coding-convention skills for Claude Code, Codex, OpenCode, Gemini CLI, GitHub Copilot, Cursor, and 50+ more agents. Each skill teaches your AI agent the idioms, patterns, and constraints of a specific library or language so you stop explaining the same things over and over.
Install one skill into your Claude Code skills directory:
# pnpm
pnpm dlx skills add avbel/ai-skills -s vitest-js
# npm
npx skills add avbel/ai-skills -s vitest-jsInstall multiple skills at once:
pnpm dlx skills add avbel/ai-skills -s rust-conventions -s anyhow-rust -s tokio-rust -s axum-rustInstall all skills from this repo:
pnpm dlx skills add avbel/ai-skills --allInstall all skills globally for a specific agent (example: Codex):
npx skills add avbel/ai-skills --global --agent codex --skill '*'Avoid combining --all with --global: --all targets every agent, including project-only Eve and PromptScript, and current skills CLI versions report those unsupported global targets as per-skill failures. This is an upstream CLI issue; the failures do not necessarily mean installs to supported agents failed. For multiple global agents, name supported agent IDs explicitly with --agent instead of --all. Install project-only agents from the project directory without --global.
List available skills:
pnpm dlx skills list avbel/ai-skillsThe installer auto-detects your agent and copies files to the right location. Pass --agent to override.
# install for Cursor
pnpm dlx skills add avbel/ai-skills -s vitest-js --agent cursor
# install for GitHub Copilot
pnpm dlx skills add avbel/ai-skills -s vitest-js --agent copilot# Install to a project-local .claude/skills directory
pnpm dlx skills add avbel/ai-skills -s vitest-js --dest .claude/skills
# Install to claude.ai project knowledge export directory
pnpm dlx skills add avbel/ai-skills -s vitest-js --dest ./exportClone the repo once, then copy the skills you need:
git clone https://github.com/avbel/ai-skills.git /tmp/ai-skills
# single skill
cp -r /tmp/ai-skills/skills/vitest-js ~/.claude/skills/
# multiple skills
cp -r /tmp/ai-skills/skills/rust-conventions \
/tmp/ai-skills/skills/anyhow-rust \
/tmp/ai-skills/skills/tokio-rust \
~/.claude/skills/- Open your project → Knowledge tab → Add
- Paste the contents of
skills/<skill-name>/SKILL.md
Or drop the file directly — claude.ai accepts markdown uploads.
| Skill | Covers |
|---|---|
anyhow-rust |
anyhow::Error, bail!/ensure!, error chaining, backtraces |
asupersync-rust |
Cx, regions, cancellation, channels, HTTP server demos |
axum-rust |
Router, handlers, extractors, State, tower middleware |
clap-rust |
Derive Parser, subcommands, env/default, shell completions |
divan-rust |
Cargo bench setup, Bencher, counters, allocation profiling |
dotenvy-rust |
.env loading, override modes, iteration, env-specific config |
eyre-rust |
Report, WrapErr, custom handlers, eyre!, anyhow migration |
futures-util-rust |
StreamExt, FutureExt, TryStreamExt, select!, join, sinks |
google-cloud-secret-manager-rust |
Secret Manager, ADC/IAM, versioned lookup |
hotpath-rust |
Rust profiling: timing, allocations, CPU, Tokio, MCP |
hud-tokio-profiler |
eBPF zero-instrumentation profiler for Tokio worker blocking; tool comparison |
moka-rust |
sync::Cache, future::Cache, TTL/TTI, eviction, weighted |
nanoprogress-rust |
Terminal progress bar conventions |
otel-observable-handles-rust |
Observable instruments, keep-alive pattern, sdk 0.27+ |
opentelemetry-rust |
Traces, metrics, logs, OTLP, propagation, sampling |
reqwest-rust |
Async client, TLS, JSON/multipart, streaming, retries |
rxrust-rust |
Observables, contexts, subjects, schedulers, async interop |
smol-rust |
block_on, spawn, Executor, Timer, Async<T>, async-compat, smol vs tokio |
tempfile-rust |
NamedTempFile, TempDir, SpooledTempFile, Builder, persist |
tokio-rust |
Patterns, anti-patterns, module reference, sync primitives, alternatives |
tokio-stream-rust |
StreamExt, ReceiverStream, StreamMap, timeout, throttle |
tonic-rust |
gRPC, protobuf codegen, streaming, interceptors, health checks |
tower-rust |
Service, Layer, timeouts, buffers, rate limits, retries |
tracing-rust |
Spans, events, #[instrument], subscriber, layers, structured logging |
walkdir-rust |
Recursive directory walk, DirEntry, filter_entry, depth control |
| Skill | Covers |
|---|---|
rust-conventions |
Rust 2024 naming, ownership, error handling, traits, iterators, concurrency |
rust-nightly |
Nightly feature gates, unstable std, -Z tooling, platforms |
rust-async-conventions |
Futures, Send/Sync, join/select, streams, pinning, cancellation |
rust-wasm-conventions |
wasm-bindgen, wasm-pack, JS interop, binary size |
design-patterns-rust |
Idioms, GoF patterns (Builder/Strategy/State/RAII), anti-patterns, principles |
rust-cookbook |
Task-to-crate recipe index: random, async, CLI, compression, db, regex, HTTP |
rust-macros |
macro_rules! fragments/repetitions/hygiene, TT munchers, proc macros (syn/quote) |
rust-high-assurance |
Static/dynamic/operational assurance, threat modeling, unsafe discipline, supply-chain, fuzzing |
rust-high-performance |
Build config, allocators, type sizes, hashers, hot-path I/O, rayon, profiling |
power-rust |
Prompt patterns that reduce LLM Rust bugs (versions, cancel-safety, SAFETY) |
| Skill | Covers |
|---|---|
biome-js |
Biome formatter, linter, XO migration, VS Code |
fastify-js |
Fastify 5 plugins, schemas, hooks, error handling, testing |
hono-js |
Routing, middleware, validation, RPC, CLI docs/search |
vitest-js |
vitest.config, vi mocking, snapshots, coverage, browser mode |
cacheable-js |
Keyv, L1/L2, wrap/memoize, hooks, distributed sync |
clickhouse-js |
MergeTree, schema, aggregates, materialized views, Node client |
close-with-grace-js |
Graceful shutdown for servers, queues, workers, CLI daemons |
duckdb-js |
SQL, complex types, window functions, Parquet/CSV I/O |
google-cloud-secret-manager-js |
Secret Manager setup, ADC/IAM, rotation, typed lookup |
opentelemetry-js |
OTel traces, metrics, SDK, auto-instrumentation, exporters |
parquet-js |
File structure, encodings, compression, hyparquet, DuckDB |
pino-js |
Log levels, child loggers, redaction, transports |
postgres-js |
Tagged templates, transactions, dynamic SQL, cursors, TypeScript |
| Skill | Covers |
|---|---|
js-conventions |
Node.js 24+, pnpm, ESM, type safety, Biome/XO/ESLint |
typescript-6 |
Existing TS 6 projects, pinned API, TS 7 blockers |
typescript-7 |
Native compiler, new projects, TypeScript 6 migration |
nodejs-26 |
Temporal, V8 14.6, Undici 8, ffi, removals |
node-rust-addon |
Rust-backed Node native addons, FFI, async |
design-patterns-ts |
GoF patterns with idiomatic TypeScript |
manual-testing-node-js |
Env setup, docker DB + migrations, SUI contracts, plan + execute |
| Skill | Covers |
|---|---|
valkey |
Features, Redis differences, Rust/Node clients, Lua, patterns/antipatterns |
clickhouse |
OLAP features, JSON/Variant types, async inserts, Rust/Node clients, antipatterns |
rocksdb |
LSM architecture, column families, compaction, Rust/Node bindings, antipatterns |
Cross-language workflow skills sharing one philosophy: cheapest sufficient process, compact readable code, justified abstractions, purposeful comments, second opinions, compounding knowledge. Start with dev-cycle.
| Skill | Covers |
|---|---|
dev-cycle |
Router and shared philosophy for the dev-* skill set |
dev-feature-lite |
Minimal path for obvious changes: implement, verify, report |
dev-feature |
Coordinated features: clarify, in-chat plan, implement |
dev-problem-solving |
Brainstorm approaches, decide, solution doc, reviewed build plan |
dev-review |
Scoped review: correctness, security, expected load, meaningful tests |
dev-testing |
Contract-focused tests, real-engine fidelity, verified API mocks |
dev-e2e-testing |
Production-parity e2e: real services, local chains, network-fault injection |
dev-debug |
Root-cause debugging, debugger configs for IDEs and CLI |
dev-knowledge |
Capture lessons in configured memory or linked project notes |
dev-code-style |
Compact, readable code with justified abstractions and sparse, purposeful comments |
| Skill | Covers |
|---|---|
code-review |
Production-focused reviews: observability, compat, migrations, idempotency, PR quality |
claude-review-code |
Delegate a local git code review to Claude Code (claude, Opus/xhigh) |
| Skill | Covers |
|---|---|
git |
Auth, merge/rebase, cleanup, executable bits, LFS, performance |
github-cli |
gh auth, repos, PRs, issues, runs, releases, API |
github-actions-ci |
pnpm/Rust/Docker CI, workflow_dispatch, caching, flow validation |
docker-compose |
compose.yaml, services, build/deploy/develop, networks, secrets |
k3s |
lightweight Kubernetes, edge, workers, storage, autoscaling, secrets |
systemd |
System and user units, timers, mounts, and transient jobs |
| Skill | Covers |
|---|---|
move-conventions |
Aptos/Sui dialect, naming, objects, capabilities, events, OTW |
sui-cli |
Network/address management, ptb, publish/upgrade, keytool |
sui-common-ops |
Common Sui queries and unsigned PTB bytes |
sui-local-dev-usdc |
Local Sui dev network, mock USDC coin, faucet, mint balances |
sui-kiosk-sdk-js |
@mysten/kiosk, KioskTransaction, listings, royalties |
sui-sdk-js |
@mysten/sui, transactions, keypairs, BCS, zkLogin |
walrus-sdk-js |
@mysten/walrus, blob reads/writes, upload relay, WASM |
Once a skill is installed, Claude Code loads it on-demand when you ask about a relevant topic. You can also invoke it explicitly:
/vitest-js
/rust-conventions
No extra configuration needed — skills are auto-discovered from ~/.claude/skills/.
See AGENTS.md for the full authoring guide, directory structure, SKILL.md format, and script requirements.
skills/
my-skill/
SKILL.md # skill definition (frontmatter + markdown)
scripts/ # optional bash scripts
bootstrap.sh
MIT