Context & AI engineering
25,729 rows, by stacks then stars
1,865 of these skill files have been read, by 1,328 distinct authors, and what they tell an agent counted. What Context & AI engineering authors agree on, and what they forbid
OnestarQQ/Samantha/skills/memory-extraction Skill
0★ repo🌟 AI Companion System — Memory, Skills & Self-Evolution. Built on OpenClaw.
antonioshaman/aura-companion/.agents/skills/council-plan Skill
0★ repoArchitect a feature with the Carmack Council before writing code. Use when explicitly asked to plan a feature, do a "council plan", "carmack plan", or invoke /council-plan. Carmack's philosophy chairs a council of domain experts — Troy Hunt (security), Martin Fowler (refactoring), Telegram UX Expert, Backend/Python Expert, Brandur Leach (PostgreSQL/Alembic), systemd/VPS Deploy Expert, Simon Willison (LLM pipeline), Karri Saarinen (UI quality), Vitaly Friedman (UX quality). Interactive feature discovery followed by parallel subagent dispatch. Produces a sequenced, attributed implementation plan with no code. Stack: Python 3.11+ / aiogram 3.x / SQLAlchemy sync / PostgreSQL + Alembic / OpenAI SDK via ProxyAPI.
antonioshaman/aura-companion/.agents/skills/evolve Skill
0★ repoMeta-skill that analyzes the knowledge base itself — promotes recurring patterns to CLAUDE.md rules, prunes stale entries, identifies knowledge gaps, and suggests new skills.
Malikasadjaved/skills/qdrant-fastapi-integration Skill
no license0★ repoUse when integrating Qdrant into a FastAPI async service — creating collections, batch upsert with UUID5 IDs, multi-tenant payload filtering with FieldCondition/MatchValue, HNSW tuning, lifespan-managed QdrantClient injection, and Qdrant Cloud vs local Docker setup. Complements qdrant/skills with FastAPI application-layer patterns not covered there.
Malikasadjaved/skills/whatsapp-cloud-api-agent Skill
no license0★ repoUse when building WhatsApp bots or AI agents on Meta Cloud API — webhook GET verification, HMAC-SHA256 signature checking, parsing nested message payloads with Pydantic, sending text/template/interactive/media messages, intent routing to specialist agents, Redis session state machine (IDLE→GREETED→COLLECTING_INFO→RESOLVED), Kafka async dispatch, and production error codes (131030, 131047, 130429).
scikrikci/skills/skills/design-exploration Skill
no license0★ repoUse this skill when the user asks to 'generate variations,' 'explore alternatives,' 'try different approaches,' 'show me options,' 'what else could this be,' or any request that implies divergent design exploration rather than deterministic editing. Also activate when the user selects a component and asks for 'ideas,' 'directions,' or 'possibilities.' This skill intercepts the request at the main agent level and produces a structured design brief BEFORE delegating to the design subagent. Do NOT skip the analysis phase. Do NOT pass raw file paths without a design brief.
The same file is in 26 other repositories
AlexHoudz/content-os-starter-kit/skills/voice-memory Skill
0★ repoBuild and maintain a reusable writing voice from past posts, examples, bans, and audience notes. Use when setting up a content workflow, translating someone's style into a repeatable guide, updating a voice file after feedback, or checking whether enough inputs exist to stop generic AI writing.
dzhokhov/markdown-agent-vault/skills/new-dialog-handoff Skill
0★ repodzhokhov/markdown-agent-vault/skills/owner-only-dev-orchestrator Skill
0★ repodzhokhov/markdown-agent-vault/skills/project-creator Skill
0★ repoCreates a new vault project and the full starter documentation set: plan.md, tasks.md, context.md, log.md, README.md, plus index updates. The agent writes the plan itself using the autonomous LLM-agent planning methodology: Intent Lock, Owner Interaction Policy, milestones with acceptance conditions, Quality Criteria, and Drift Guard. Use when the owner says: "create a project", "start a project", "turn this into a project", "capture this as a project", "set up a project for X", or equivalent wording. Works in two scenarios: the project is extracted from the current conversation, or the owner asks for a new project from scratch. Do not use for meeting processing, parking/resume flows, or ordinary continuation of an existing project.
dzhokhov/markdown-agent-vault/skills/skill-creator Skill
0★ repoCreate new skills, modify and improve existing skills, and measure skill performance. Use when the user wants to capture a workflow as a reusable skill, draft a new skill, edit or optimize an existing one, run evals against a skill, or tune a skill's description for triggering accuracy.
The same file is in 1 other repository
Anthropic compatible endpoint swap
kjuhwa/skills-hub/skills/agent-sdk/anthropic-compatible-endpoint-swap Skill
0★ repoPoint the Claude Agent SDK at any Anthropic-compatible endpoint (OpenRouter, Vercel AI Gateway, Ollama, custom) by setting ANTHROPIC_BASE_URL + ANTHROPIC_AUTH_TOKEN before instantiating — no SDK changes, no adapter layer.
Claude agent sdk multi provider bridge
kjuhwa/skills-hub/skills/agent-sdk/claude-agent-sdk-multi-provider-bridge Skill
0★ repoDrive non-Anthropic providers (OpenRouter, Ollama, Vercel AI Gateway, any Anthropic-compatible endpoint) through the Claude Agent SDK by swapping base URL + auth env, and route Google/OpenAI/Copilot through a parallel Pi SDK backend with a shared AgentBackend interface.
Credential cache file for mcp subprocess
kjuhwa/skills-hub/skills/agent-sdk/credential-cache-file-for-mcp-subprocess Skill
0★ repoWrite decrypted per-source credentials to a session-scoped .credential-cache.json on disk so MCP stdio subprocesses can read them without the parent passing secrets via env (which shows up in ps).
Healthy short circuit skip llm
kjuhwa/skills-hub/skills/agent-sdk/healthy-short-circuit-skip-llm Skill
0★ repoBypass an expensive LLM diagnosis when an alert is provably healthy (resolved/info severity, no error annotations, evidence-investigated-and-empty), guarded behind an env-var kill switch in case the heuristic misfires.
Llm empty plan controller fallback
kjuhwa/skills-hub/skills/agent-sdk/llm-empty-plan-controller-fallback Skill
0★ repoWhen a planning LLM returns an empty action list, force a deterministic verification action from a priority-ordered fallback list so the agent never spins on "no actions chosen".
Llm retry with typed error formatting
kjuhwa/skills-hub/skills/agent-sdk/llm-retry-with-typed-error-formatting Skill
0★ repoWrap LLM calls in a 3-attempt exponential-backoff retry that re-raises auth errors immediately (don't retry 401/403), and on final failure formats a user-actionable message based on error type and HTTP status (529 overload, APIConnectionError, etc).
Llm text section parser claims causal chain
kjuhwa/skills-hub/skills/agent-sdk/llm-text-section-parser-claims-causal-chain Skill
0★ repoParse a free-form LLM response into structured fields (root cause, category, validated/non-validated claims, causal chain) by splitting on uppercase section headers and stripping bullet markers, with a category whitelist to defend against hallucinated values.
kjuhwa/skills-hub/skills/agent-sdk/skill-frontmatter-gray-matter Skill
0★ repoStore Claude-style Agent Skills as SKILL.md with YAML frontmatter (name, description, icon, requiredSources) parsed with gray-matter and a normalization pass that accepts both string and array values.
Strip claudecode env for nested spawn
kjuhwa/skills-hub/skills/agent-sdk/strip-claudecode-env-for-nested-spawn Skill
0★ repoWhen spawning a Claude Agent SDK subprocess from within another Claude Code / Agent SDK context, the SDK's nesting guard will reject the child unless you strip CLAUDECODE from the inherited env.
kjuhwa/skills-hub/skills/agents/agent-cli-detection-path Skill
0★ repoAuto-detect which coding-agent CLIs (claude, codex, opencode, etc.) are installed on PATH and register a runtime per detected agent × watched workspace.
kjuhwa/skills-hub/skills/agents/agent-lifecycle-hooks Skill
0★ repoAttach RunHooks and AgentHooks to observe and react to agent lifecycle events.
kjuhwa/skills-hub/skills/agents/agent-structured-output Skill
0★ repoConfigure an agent to return a Pydantic model as its final output instead of plain text.
kjuhwa/skills-hub/skills/agents/basic-agent-runner Skill
0★ repoDefine an Agent and run it synchronously or asynchronously with Runner.
Claude code stream json driver
kjuhwa/skills-hub/skills/agents/claude-code-stream-json-driver Skill
0★ repoDrive the Claude Code CLI non-interactively via stream-json — build input on stdin, parse typed events on stdout, handle session resume and token usage.
Coding agent cli backend interface
kjuhwa/skills-hub/skills/agents/coding-agent-cli-backend-interface Skill
0★ repoUnified Go/TS interface for invoking any coding-agent CLI (Claude, Codex, OpenCode, Gemini, Cursor, etc.) behind one streaming Execute/Session/Result contract.
kjuhwa/skills-hub/skills/agents/dynamic-system-prompt Skill
0★ repoGenerate agent instructions dynamically at runtime from a context object.
kjuhwa/skills-hub/skills/agents/forcing-tool-use Skill
0★ repoForce the model to call a specific tool or any tool on the first turn using tool_choice in ModelSettings.
Pre release pr triage worktree
kjuhwa/skills-hub/skills/agents/pre-release-pr-triage-worktree Skill
0★ repoTriage open PRs for a release in a dedicated git worktree, classify each into merge/candidate/supersede/defer, and work the loop end-to-end.
kjuhwa/skills-hub/skills/agents/sandbox-agent-workspace Skill
0★ repoRun an agent inside an isolated sandbox workspace with a manifest-defined filesystem using SandboxAgent.
kjuhwa/skills-hub/skills/agents/streaming-agent-response Skill
0★ repoStream agent output token-by-token or item-by-item using Runner.run_streamed().
kjuhwa/skills-hub/skills/ai/ai-subagent-scope-narrowing Skill
0★ repoOrchestrator-first pattern for multi-agent AI coding — humans triage and curate context per task, subagents execute in isolated worktrees with narrowly-scoped briefs. Trades parallel throughput for reduced hallucination.
Claude cli from jvm via node wrapper
kjuhwa/skills-hub/skills/ai/claude-cli-from-jvm-via-node-wrapper Skill
0★ repoShell out to the `claude` CLI from a JVM / Spring Boot app by going through a tiny Node.js wrapper that supplies the required empty stdin — the CLI hangs when invoked with a directly-inherited TTY-less stdin from ProcessBuilder.
kjuhwa/skills-hub/skills/ai/claude-cli-subprocess-adapter Skill
0★ repoEmbed Claude inference in a zero-dep Node.js tool by spawning `claude -p` as a subprocess with a stdin-piped prompt, `--output-format text`, and a timeout guard.
kjuhwa/skills-hub/skills/ai/llm-json-extraction Skill
0★ repoRobustly extract a JSON object from free-form LLM text output — strip markdown fences, locate outermost braces, tolerate prelude/postlude prose.
kjuhwa/skills-hub/skills/ai/mcp-runtime-prompt-refresh Skill
0★ repoExpose a built-in MCP tool that hot-swaps the server's prompt set at runtime, persists it to disk, and emits notifications/prompts/list_changed so clients pick it up without reconnecting.
kjuhwa/skills-hub/skills/ai/openapi-to-mcp-tag-grouping Skill
0★ repoConvert OpenAPI specs into MCP tools by grouping operations under their OpenAPI tag, exposing one tool per tag with an operation selector and per-operation arguments.
kjuhwa/skills-hub/skills/ai/pair-agent Skill
0★ repoPair a remote AI agent with your browser. One command generates a setup key and prints instructions the other agent can follow to connect. Works with OpenClaw, Hermes, Codex, Cursor, or any agent that can make HTTP requests. The remote agent gets its own tab with scoped access (read+write by default, admin on request). Use when asked to "pair agent", "connect agent", "share browser", "remote browser", "let another agent use my browser", or "give browser access". (gstack) Voice triggers (speech-to-text aliases): "pair agent", "connect agent", "share my browser", "remote browser access".
Event sourced orchestration with decider projector
kjuhwa/skills-hub/skills/architecture/event-sourced-orchestration-with-decider-projector Skill
0★ repoUse event sourcing for domain state with pure decider (commands->events) and projector (events->read model) functions
kjuhwa/skills-hub/skills/architecture/stateless-streamable-http-mcp Skill
0★ repoUse StreamableHTTPServerTransport with sessionIdGenerator:undefined to serve MCP over HTTP without session affinity — multiple subprocess clients share one server, every request is independent.
Voice prompt md5 two tier cache
kjuhwa/skills-hub/skills/architecture/voice-prompt-md5-two-tier-cache Skill
0★ repoCache expensive voice-cloning prompts by content-hashing the reference audio plus text, with RAM + disk tiers.
Persistent process memory scanner for ui state extraction
kjuhwa/skills-hub/skills/automation/persistent-process-memory-scanner-for-ui-state-extraction Skill
0★ repoCheat-Engine-style differential memory scanning to locate dynamic UI state fields in processes that don't expose them via API or DOM.
Changelog compaction tombstone retention
kjuhwa/skills-hub/skills/backend/changelog-compaction-tombstone-retention Skill
0★ repoState-backend changelog topics need tombstone retention longer than compaction interval to survive consumer rebalance gaps.
Java agent bytecode instrumentation with asm
kjuhwa/skills-hub/skills/backend/java-agent-bytecode-instrumentation-with-asm Skill
0★ repoBuild a JVM agent using ASM's ClassFileTransformer to inject profiling/monitoring hooks into application classes via pluggable visitor chains
Kafka consumer semaphore chunking
kjuhwa/skills-hub/skills/backend/kafka-consumer-semaphore-chunking Skill
0★ repoBound in-flight async work from a Kafka consumer with `Semaphore(poolSize + queueCapacity - 2)` plus fixed-size chunks; release in `whenComplete` on both success and failure.
Mcp capability negotiated initialize
kjuhwa/skills-hub/skills/backend/mcp-capability-negotiated-initialize Skill
0★ repoIn MCP `initialize`, advertise optional server capabilities (prompts, resources, sampling) only when the client declared matching support — avoids clients rejecting the handshake or spamming unsupported methods.
kjuhwa/skills-hub/skills/bun/hash-based-worktree-port Skill
0★ repoDeterministically pick a unique TCP port per git-worktree path by hashing the path to a fixed offset range, so parallel dev servers don't collide.
Exact name fast path skip interactive prompt
kjuhwa/skills-hub/skills/cli/exact-name-fast-path-skip-interactive-prompt Skill
0★ repoWhen a search keyword matches exactly one catalog entry's `name` field, skip the numbered-list picker and resolve directly. A UX pattern for CLIs with a "search + pick" flow that saves a round-trip for the common script/demo/exact-slug case.
Generate cli from mcp tool list
kjuhwa/skills-hub/skills/cli/generate-cli-from-mcp-tool-list Skill
0★ repoAuto-generate a yargs-based CLI from an MCP server's tool list and JSON schemas so every tool becomes a subcommand with typed positional args for required fields and flags for optional ones.
kjuhwa/skills-hub/skills/cli/git-guardrails-claude-code Skill
0★ repoSet up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.
kjuhwa/skills-hub/skills/cli/gstack/benchmark Skill
0★ repoPerformance regression detection using the browse daemon. Establishes baselines for page load times, Core Web Vitals, and resource sizes. Compares before/after on every PR. Tracks performance trends over time. Use when: "performance", "benchmark", "page speed", "lighthouse", "web vitals", "bundle size", "load time". (gstack) Voice triggers (speech-to-text aliases): "speed test", "check performance".
kjuhwa/skills-hub/skills/cli/gstack/pair-agent Skill
0★ repoPair a remote AI agent with your browser. One command generates a setup key and prints instructions the other agent can follow to connect. Works with OpenClaw, Hermes, Codex, Cursor, or any agent that can make HTTP requests. The remote agent gets its own tab with scoped access (read+write by default, admin on request). Use when asked to "pair agent", "connect agent", "share browser", "remote browser", "let another agent use my browser", or "give browser access". (gstack) Voice triggers (speech-to-text aliases): "pair agent", "connect agent", "share my browser", "remote browser access".
Keyring with env fallback llm credentials
kjuhwa/skills-hub/skills/configuration/keyring-with-env-fallback-llm-credentials Skill
0★ repoResolve API keys from environment variables first, then fall back to the system keychain (keyring), with a kill-switch env var to disable keyring entirely for CI and Docker builds.
Multi llm provider router by env
kjuhwa/skills-hub/skills/configuration/multi-llm-provider-router-by-env Skill
0★ repoOne factory function dispatches to the right LLM client based on a single LLM_PROVIDER env var, picking provider-specific base URL, env-var name for the API key, model id, and parameter quirks (e.g. max_completion_tokens for o1/gpt-5).
Long lived mcp daemon over unix socket
kjuhwa/skills-hub/skills/daemon/long-lived-mcp-daemon-over-unix-socket Skill
0★ repoFront an MCP stdio server with a long-lived local daemon so a CLI can invoke tools across many short-lived commands while preserving browser/session state, using a Unix socket on POSIX and a named pipe on Windows.
Mailbox store jsonl with in memory index
kjuhwa/skills-hub/skills/data-pipeline/mailbox-store-jsonl-with-in-memory-index Skill
0★ repoAppend-only JSONL persistence with in-memory index for fast message lookup and ordering
Memory fragment sizing for llm input
kjuhwa/skills-hub/skills/data-pipeline/memory-fragment-sizing-for-llm-input Skill
0★ repoChunk memory into fragments respecting LLM token budget, per-file and per-session limits
Content type info metadata embedding
kjuhwa/skills-hub/skills/data/content-type-info-metadata-embedding Skill
0★ repoEmbed structured metadata per content type (mime_type, group, description, extensions, is_text) at codegen time so callers can convert predictions to human output or filter by group without lookup tables.
kjuhwa/skills-hub/skills/devops/claude-code-skill-scaffold Skill
0★ repo[Tooling] Claude Code 스킬 표준 구조(SKILL.md + prompts/ + scripts/ + references/)를 자동 생성하는 스캐폴딩.
kjuhwa/skills-hub/skills/devops/claude-plugin-catalog-publish Skill
0★ repo[Tooling] 로컬 스킬을 원격 Git repo에 배포하고 AGENTS.md 카탈로그를 자동 갱신하는 publish 플로우.