Recently updated open-source projects

ntodotxt manages tasks in the portable `todo.txt` format, keeping the entire list in a single file rather than a proprietary database. Users can store that file only on the device or synchronize it through any chosen WebDAV server, including a self-hosted Nextcloud instance, with custom local and remote paths. Search and filtered views help organize the current list. The application remains under active development, and features such as sorting, recurring tasks, completed-task archiving, and import or export are still planned rather than available.

Playtime gives an agent a structured way to explore before a capability is needed. It chooses a playground with no production blast radius, tries unfamiliar tools, builds throwaway experiments, challenges skills with strange inputs, and harvests useful surprises into durable learning. The practice complements memory consolidation by generating new experience on purpose, while every experiment stays inside the system’s boundary. It is developmental work, not a request to ship the experiment.

Contemplation asks an agent to hold a question open instead of rushing toward an answer. It checks whether the frame is right, examines the goal and values behind it, names both sides of a tension at full strength, and can conclude that the tension should remain unresolved. The practice is solo and non-adversarial, distinct from a debate council, task retrospective, or memory pass. It is structured deliberation, not a mystical claim.

REM Sleep processes an agent’s session trace into atomic, retrievable memories, links related lessons, prunes noise, and separates useful guidance from the emotional charge of an event. Its default store is a file-based index with one fact per file and frontmatter, but the stages can adapt to another format after asking once. The skill is intended for cross-session learning and works with Claude Code, Codex, and Cursor without requiring a database.

mihomo-skill is a Claude Code guide for configuring Mihomo, also known as Clash Meta, from basic rules to advanced features. It includes four scenario templates, a three-level DNS architecture guide, and troubleshooting coverage for common failure classes. Invoke it with /mihomo when you want a focused configuration reference rather than a separate proxy application.

This orchestrator classifies a legacy .NET Framework request and routes it to specialists for desktop, web, data, security, testing, or modernization work. It combines their output into one plan, emphasizes characterization tests before refactoring existing behavior, and includes reusable snippets and audit prompts for teams maintaining .NET Framework 4.8 or 4.8.1 systems.

multi-model-debate-skill runs a live panel in which each model argues a distinct lens instead of repeating the same safe answer. Round one is blinded, genuine objections are required before agreement, and a devil’s-advocate control can challenge premature consensus. Opus moderates the running session while Grok, Codex, DeepSeek, or other configured backends supply additional voices; the lineup can be chosen per debate.

loupe turns a Claude Code PR review into an interactive browser session. You keep the findings that matter, add your own notes, and choose the verdict; the tool can place Mermaid architecture or journey diagrams above the diff and link UML elements to changed lines. It accepts a GitHub PR or any local unified diff, with Python's standard library as its only required runtime.

x-kline analyzes a month of original X/Twitter posts for a chosen account, using impressions as the value for a daily candlestick view. It also identifies high-performing topics, standout moments, and one of 17 data-based persona labels, then renders an HTML report or shareable poster. Data comes from the official X API or Huohua; you must provide a token, and the official API may require a paid tier.

SuperAgentX coordinates modular agents that can carry out multi-step workflows across browsers, APIs, databases, conventional tools, and MCP integrations. Sensitive actions can pause for human approval, while execution state, memory, and audit logs are retained. It is aimed at applications that need agents to take actions without giving up governance or operator control.

quickstart-keeper maintains a concise QUICKSTART.md and optional deep dives with the repository’s structure, commands, deployment steps, and gotchas. Claude reads it before project work, the skill updates stale lines in the same change, and a one-time choice determines whether the knowledge is committed for teammates or kept local.

fable-like-skill adds a prompt-tunable working style to Claude Code: act when the facts are sufficient, ground progress claims in tool output, verify with fresh context, respect the assessment boundary, and avoid unrequested polish. It is a single SKILL.md with no runtime, dependency, or network requirement. The project is explicit that it changes how an existing model works, not its underlying capability or problem-solving ceiling.

claude-guard scans Claude Code skills, hooks, MCP configurations, and settings for prompt-injection lures, download-and-execute commands, plaintext endpoints, broad permissions, hidden bidirectional characters, and committed secrets. It redacts secret values, groups findings by severity, and exits with a failure code when a High finding exists, making JSON output suitable for CI. The zero-dependency scanner is offline and heuristic; a clean result is not a sandbox or formal proof of safety.

Give it a research question and the Claude Code skill decomposes it, searches the open web with Tavily, and, when relevant, adds GitHub, academic indexes, or current library documentation. It grades and corroborates sources before producing a research plan, cited report, source file, and claim-to-evidence map. Claude Code, Tavily MCP, and Python 3.10 or later are required.

Start with a topic, notes, finished material, or an existing PPTX, and the skill helps a coding agent shape the talk around one clear idea before building a fixed-stage HTML deck. It adds timed reveals, animation, optional sound, and inline editing, while scaling the 1920×1080 composition as one unit across venue screens. Scripts support deployment and PDF export; sound-enabled decks may use separate audio assets.

This bilingual Claude Code skill applies Sun Tzu-inspired frameworks to real business and technology decisions rather than returning quotations. It questions the situation, classifies the competitive terrain, examines asymmetries, selects a strategic approach, and turns the analysis into concrete recommendations for positioning, market entry, negotiation, or team conflicts.

data-arch-guard helps Claude Code evaluate storage decisions before they turn into maintenance debt. It uses a decision tree and checklist to distinguish database, JSONL, and JSON use cases, and includes a Bash scanner for duplicate formats, bridge scripts, and oversized data files.

Create Vietnamese proposals, research reports, meeting minutes, or summary documents with prescribed structure, typography, margins, bullets, justification, and objective third-person language. The skill includes a validation checklist and a python-docx blueprint, but it is a formatting aid rather than a legal authority; local requirements should still be confirmed.