Claude Skills by agentskillexchange

Use Darwin Skill to evaluate, improve, test, and keep or roll back changes to a single agent SKILL.md with a validation-gated, human-reviewed optimization loop.

Use Jumbo Context when coding agents need local goal state, project memory, session continuity, and context packets that travel across Claude, Codex, Copilot, and other harnesses.

Paseo runs a local daemon for Claude Code, Codex, Copilot, OpenCode, and Pi agents so operators can start, monitor, attach to, and hand off coding-agent work from desktop, mobile, web, or CLI clients.

Set up OpenHuman as a local-first personal agent harness that syncs context, builds durable memory, and coordinates multi-agent workflows behind approval gates.

Use Maestro Flow when a coding operator wants intent-driven planning, execution, verification, review, retry, and long-running multi-agent loops from one workflow surface.

Use Kandev when an operator needs to plan, run, isolate, review, and ship multiple coding-agent tasks through a controlled kanban and pipeline workspace.

Use Wmux to fan out Claude Code, Codex, Gemini, and other agent CLIs across native Windows panes, isolated worktrees, approval gates, durable channels, and browser/MCP automation.

Use Microsoft's POML to turn advanced prompts into structured, reusable prompt assets with markup, data components, templating, styling, SDK rendering, and VS Code testing.

Checks official cloud and SaaS status evidence through the OutageDeck MCP server or anonymous REST API before an agent changes code for 5xx, timeout, deployment, authentication, or integration failures.

Run MonkeyOCR over PDFs, scanned pages, formulas, and tables to produce structured Markdown/text that downstream agents can ingest and reason over.

Extract document text, tables, logical structure, and metadata into normalized output before RAG, review, or workflow automation.

Use kordoc when an agent needs to parse HWP, HWPX, PDF, Excel, or DOCX files into Markdown, fill forms, compare documents, or expose document tools through MCP.

Run OmniParse locally or as an API service to turn documents, images, audio, video, and web pages into structured Markdown-ready inputs for agent and RAG workflows.

Use dots.ocr when an agent needs repeatable multilingual document layout parsing before retrieval, extraction, review, or downstream automation.

Install a Claude Code skill that learns a user's social voice, builds a 14-day content calendar, drafts posts, and publishes only after explicit confirmation.

Structured self-review checklist to run before committing substantial code changes. Covers edge cases, error paths, test coverage, documentation, and code quality — so CI is a last resort, not your first reviewer.

Use PyMuPDF to extract text, layout metadata, tables, images, Markdown, or JSON from PDFs and documents before an agent ingests or reviews them.

Use OpenBidKit Yibiao to turn tender materials, company knowledge, and model-provider settings into reviewable bid proposal drafts with reuse and risk-check workspaces.

Use ContribAI when an agent should analyze a repository, respect maintainer consent, and prepare bounded draft contribution proposals with evidence receipts.

Use itr-wala when an agent needs a review-first workflow for reading Indian tax documents, running deterministic AY 2026-27 tax calculations, comparing regimes, and guiding the user up to the pay, submit, and e-verify steps.

Generate dense, sparse, image, and reranking embeddings locally before writing vectors into Qdrant or another retrieval stack for agent memory and RAG workflows.

Gives coding agents a repeatable file-backed planning workflow using task_plan.md, findings.md, and progress.md so long-running work survives /clear, compaction, crashes, and handoffs.

Use video-recap-skills when an agent should turn source video into a Chinese narration recap with scene understanding, scripting, voiceover, subtitles, and an optional editable JianYing draft.

Turns Claude Code, Codex, or another skills-aware coding agent into a Remotion-based product video operator with shot cards, motion previews, templates, capture scripts, sound design, and visual QA.

Use Scalene when a Python agent, tool server, or retrieval pipeline needs line-level CPU, GPU, and memory profiling before targeted optimization.

Restructure large documentation files (500+ lines, 5k+ tokens) into slim indexes with on-demand detail directories. Reduces always-included context by 40-60% while preserving full accessibility. Ideal for documentation auto-included in every agent context window.

Answers short pronunciation questions about developer tools, AI models, acronyms, and project names by using the say-it CLI to play local audio and return IPA, a plain-English respelling, source evidence, and contested alternatives.

Uses the Proofed CLI and current-subject completion receipts to reject unsupported coding-agent completion claims, rerun repository-configured tests, and detect stale PASS evidence after code changes.

Use SkillForge to triage, create, test, review, compile, and maintain Claude Code and Codex-ready skills with baseline failures and regression evals.

Use ClickHouse's official MCP server when an assistant needs schema-aware, reviewable analytics access to ClickHouse through bounded tools instead of ad hoc database credentials.

Use DBHub to expose guarded, token-efficient database inspection and SQL tools to MCP clients across Postgres, MySQL, SQL Server, MariaDB, and SQLite.

Give MCP-capable agents a live global-intelligence surface for country briefs, risk scores, conflict, cyber, market, weather, and infrastructure monitoring.

Bridge Cursor, Claude Code, and other MCP clients into Figma so agents can inspect selections, create nodes, annotate designs, and apply bulk edits.

Use website-rebuild-skill when an agent needs to classify a public site, capture a read-only evidence mirror, reconstruct behavior from shipped assets, and verify the rebuild with console, network, DOM, geometry, and pixel-diff gates.

Use OpenAdapt to demonstrate a repeated browser or GUI workflow once, compile it into a deterministic replay bundle, and certify or halt runs with reviewable reports when interfaces drift.

Wrap Claude Code, Codex, Gemini CLI, OpenClaw, or another agent command with AgentSight to capture processes, files, network destinations, prompts, and reports.

Use OMNI to distill noisy terminal output before coding agents read it while preserving recoverable raw logs for exact follow-up inspection.

Use Honey to make coding agents emit less code, less prose, and denser agent-to-agent handoffs while preserving correctness boundaries.

Build an MCP Apps resource, attach it to a tool through _meta.ui.resourceUri, and verify the host renders the interactive UI alongside tool results.

Use Sepia to write, review, minimally refactor, or fully recreate fiction and professional prose with agent-native de-AI writing protocols.

Use jDocMunch MCP when an agent needs to index local or repository documentation, search by heading-level structure, and retrieve only the exact section needed instead of flooding context with full files.

Use Hunk to keep a live terminal review UI open for agent-authored code changes, with Git, Jujutsu, Sapling, patch input, watch mode, and inline agent annotations.

Use Claude Legal Skill to guide an agent through contract risk review, CUAD-style clause checks, market benchmarks, and lawyer-ready redline suggestions.

Use repository-settings when an operator wants GitHub repository configuration changes reviewed as code before branch protection, labels, teams, or other settings are applied.

Use Shepherd Model Gateway to route OpenAI, Anthropic, Responses API, embeddings, and MCP tool traffic across self-hosted and cloud model backends with cache-aware policies and observability.

Use cc-thinking-skills to install and invoke a portable catalog of mental-model and critical-thinking skills for decisions, debugging, systems analysis, risk, and strategy across compatible coding agents.

Use Plano when an agent application needs an out-of-process data plane for multi-agent routing, model agility, guardrails, traces, and OpenAI-compatible traffic instead of custom glue in every service.

Give agents a routing pack for authorized APK, binary, mobile, API, malware, CTF, and penetration-test analysis with tool checks and case structure.

Give agents a bilingual marketing SOP system for plans, copy, audits, content calendars, campaign briefs, design prompts, and role-based workflows.

Run a local proxy that lets Codex and Claude Code use configured LLM providers, routed models, and account pools while keeping the normal coding-agent workflow.