Recently updated open-source projects

The Claude Code skill provides an offline, modular reference for authentication, order placement, market data, WebSocket subscriptions, CTF operations, bridging, gasless transactions, fees, and error handling. Examples cover both Python and TypeScript. It guides integration work rather than operating as a trading bot, and real deployments must respect current geoblocking and rate-limit rules.

This skill teaches Claude Desktop or Claude Code when and how to call the VirtualSMS MCP for phone verification. An agent can search availability, buy a number, wait for an SMS code by WebSocket or polling, swap a failed number, and cancel or refund unused orders across thousands of services and many countries; access requires the MCP connection and a VirtualSMS API key.

OpenClaw Auditor gives Claude Code a static review workflow for proposed OpenClaw configuration changes. It is designed to consult the current documentation index, capture a configuration snapshot, compare schema expectations, score risk, and relate findings to documented failure modes before anything is applied. The skill can format results for Notion and requires curl and jq, with the OpenClaw CLI optional.

Use this system to structure a learning topic into knowledge blocks with conditions, outcomes, examples, and fast decision patterns. It schedules practice repetitions, focus blocks, and daily checkpoints, then tracks scores, streaks, pass rates, and progress reports. Claude Code commands can create a seven-day plan or start interactive practice, while the local Python tools store the learning data without external packages.

Use a CLI to read and manage Vapi assistants, squads, tools, calls, phone numbers, and files through the REST API, preserving large system prompts and transcripts that the Vapi MCP server cannot return intact. It emits JSON to stdout or a file for local jq workflows and includes a Claude Code skill for AI-assisted operations. Vapi authentication is required.

This Claude Skill turns supplied notes, text, files, or ideas into academic Word documents or PowerPoint decks for Tec de Monterrey. It applies the institution's blue palette, Arial typography, logo treatment, academic cover pages, headers, footers, and presentation layouts, with examples for individual work, team reports, and project defenses. It is aimed at Tec students and faculty; the included institutional logos are intended only for that context.

Maat audits strategy reports, M&A memos, market analyses, RAG knowledge bases, prompt libraries, and other business artifacts shaped by AI. It checks whether strategic claims can be traced to evidence, whether uncertainty was lost, whether authority is being faked, and whether counter-analysis is missing. The result includes an executive verdict, principle-by-principle findings, risky claims, a reconstructed prompt, and a machine-readable 0–10 score.

Provide a screenshot, PDF, HTML snippet, or form description in Claude Code, Cursor, or Windsurf, and the tool returns validated FormEngine JSON plus a runnable App.tsx. The schema can round-trip through the online visual builder and render with RSuite, Material UI, or Mantine, making it easier to edit and reuse than one-off JSX. It currently depends on an LLM host that can load the skill or rule; the standalone MCP server is still planned.

copywriting-15-patterns turns a Japanese headline or social-hook request into grouped options drawn from 15 patterns across five functional clusters. You can ask for a specific pattern, a cluster, or the default multi-pattern mix, with the skill producing several candidates rather than one rewrite. It avoids invented numbers, fake deadlines, and deceptive claims, and runs as a Claude Code skill.

Oscar adds one of five politeness levels to Claude Code prompts, with an optional matching reply style or clean output. It also includes a benchmark scaffold for testing the cited paper’s method, while clearly noting that the original evidence came from a small GPT-4o multiple-choice study and may not generalize.

Route each request through a small trigger taxonomy so substantial feature or debugging work loads the appropriate skill, while lookups and trivial edits proceed directly without forced narration. Skill Radar also respects terse modes and restates instruction priority when dispatching subagents. It is designed as a lean triage alternative for Claude Code users who want fewer false-positive skill invocations.

Tianluo teaches an agent how to run long, cron-driven work without repeatedly asking for routine decisions. It enumerates decision forks before launch, stores machine state separately, diagnoses failures across five layers, and applies bounded retries with an escalation ladder. The project is a methodology skill rather than a scheduler or runtime, so you still provide the storage, trigger system, and task-specific implementation; it is not intended for short or judgment-heavy work.

Clawproof Checks evaluates an agent system across ten operational areas, including least privilege, audit trails, prompt injection, human escalation, rollback, secrets, and regression testing. The packaged Claude skill can inspect a codebase, runbook, or configuration and return a scored report with prioritized remediation. The same checks are also available as YAML and consolidated JSON for vendor-neutral programmatic use.

session-bug-hunter scans conversation history and changed files for errors, unfinished TODOs, security concerns, and other bug signals. It groups findings by severity and type, shows the full list for approval, then opens structured GitHub issues with locations, reproduction steps, and suggested fixes. It requires Claude Code and an authenticated GitHub CLI.

Use Claude Code to list, search, create, update, and close Redmine issues through the required Redmine MCP server. The skill can add polished English notes, track individual or bulk time entries, discover project metadata and custom fields, and fill leftover hours against a configured management issue. It needs an authenticated MCP server and a suitably scoped Redmine API key.

Use dedicated modes to decompose automotive requirements, check INCOSE and EARS quality, review C code against MISRA-C, or examine ADAS levels, ASIL, AUTOSAR, and SOTIF concerns. The skill works in several AI tools that accept system instructions and is intended for early analysis, training, and prototyping. It is not a certified tool and cannot replace formal static analysis, traceability systems, or independent safety assessment.

Perspectives first gathers context, then creates four to six personas tailored to the decision, product, design, or code under review. Its default council gives independent opinions and a synthesis; deeper tiers add structured frameworks or dispatch sub-agents for code, security, usability, documentation, research, and performance work when explicitly requested or when serious concerns emerge.

Turn Claude Code into a chapter-aware reading companion. It reads scenes and quotes from a configured plaintext library instead of relying on recalled plot details, keeps discussion within your current chapter, and records reactions and predictions in a per-book journal. For a series, completed books remain available for cross-book callbacks without exposing material beyond your reading progress.