Recently updated open-source projects

im-designer translates subjective design language into requests that specify the user problem, target component, state, visual property, scope, and accessibility considerations. When a phrase has multiple plausible meanings, it can render HTML options for comparison before producing versions for designers, developers, AI design tools, or Figma and Penpot, while keeping color choices token-based.

axion-skill routes a forecasting question through research agents that search the web, inspect market data and SEC filings, and run Fermi estimates before returning a probability. Each forecast includes a 0–1 estimate, interval, resolution date, and cited basis. It requires an Axion API key with prepaid credits, so the output is a research forecast rather than a guarantee or a substitute for domain judgment.

Text Message Video Skills renders a scripted conversation or CSV as an animated iMessage, WhatsApp, or SMS-style video. Bubbles appear one at a time, typing indicators pulse, send sounds land with each message, and the thread scrolls as the exchange grows. The same template can render many conversations, making it useful for short-form story content. The skill handles presentation and timing; the author remains responsible for the script and its platform context.

nvc-communication turns complaints, refusals, criticism, apologies, and anger into more usable language using the observation, feeling, need, and request framework. It supports both expressing yourself and empathizing with the other person, with focused paths for requests, boundaries, gratitude, and self-care. The output stays conversational rather than formulaic.

Coro Code is a Rust terminal coding agent with an interactive UI plus built-in shell and file tools. It runs on macOS, Linux, and Windows and currently centers on OpenAI or OpenAI-compatible endpoints. Native Anthropic and Google support, tool-call permissions, and sandboxing are still listed on the roadmap, so users must set their own execution boundaries for now.

Map an existing business process by having the agent interview you about actors, sequence, tools, decisions, and pain points before it draws anything. Process Builder converts the confirmed model into an editable draw.io swimlane and can propose quick-win or fuller automation variants for a future-state process. Its zero-dependency Python generator validates the JSON and handles layout; the interview guidance is currently written in Italian.

Dev Fable helps Claude Code handle multi-step development by planning phases, routing work to specialized subagents, and synthesizing their results. It can run independent perspectives in parallel for high-stakes decisions, while separate roles cover architecture and debugging, repetitive edits, and read-only exploration. The bundled agents are defaults that teams can remap to their own models and tools.

Nori is a native local CLI that puts Claude, Codex, Gemini, and other ACP-compatible agents behind one interactive terminal UI. It can reuse existing authenticated sessions on the machine, and the /agent command switches providers, making it useful for comparing tools without changing the surrounding workflow.

Joplin Note Tabs adds a tab panel to the Joplin desktop app so you can keep several notes open and pin important ones. Tabs can include navigation controls, checklist progress, and breadcrumbs, and they can be arranged horizontally or vertically and controlled with shortcuts. Pinned tabs persist in the local database but do not sync to other devices, and the plugin requires Joplin 1.8.2 or newer.

Mac Vibe Triage investigates memory pressure, battery drain, orphaned development processes, and browser helpers left behind by AI-assisted work. It gathers read-only evidence across CPU, cumulative time, RSS, process ancestry, settings, and installation locations before classifying candidate PIDs. A kill requires your explicit approval in the current session, and the safety checks reject root, system, other-user, and active-agent processes.

rednote-writing-skill turns links, screenshots, notes, or rough ideas into finished Xiaohongshu posts, and can also critique an existing draft. Collage, Co-create, and Review modes preserve the author’s judgment and voice without inventing facts. After the body is written, a mechanical pass checks punctuation, automated-writing tells, and platform-risk terms, followed by a fresh human-sounding pass; titles and cover lines come last. The voice is tuned to one account through its real post examples, so another account needs new examples and positioning.

Riichi Mahjong Calculator analyzes hand efficiency, open-hand choices, winning hands, and scores from han and fu. Its default rules follow Mahjong Soul and Tenhou conventions with some customization. Tile inputs can be recognized from images or the clipboard, with the recognition model tuned for Mahjong Soul screenshots. Calculations run locally, and builds are available for web, Android, desktop, and self-signed iOS installation.

Travel Planner collects missing trip details in one round, then previews the itinerary, budget, risks, and every item it plans to write. After confirmation, it creates a linked Apple Notes itinerary and checklist, Calendar events for fixed times, Reminders for pre-departure tasks, and Apple Maps links. Writing to Apple apps requires macOS and the relevant AppleScript automation permissions.

Performance Report Assistant interviews the user, gathers notes and git history, inspects an Excel template, and proposes the exact cells or sections to change before filling a copy. It supports weekly updates, reviews, promotion narratives, and customer reports while preserving workbook structure. Complex mappings may still need confirmation, and the WeCom collector is Windows-specific.

Loom is an experimental Rust coding-agent research project with a REPL, file and code tools, server-side model proxying, conversation persistence, and remote execution components. Its repository explicitly restricts use to Geoffrey Huntley and warns that the software is unstable, unsupported, and subject to breaking API changes.

Work Backstory records the causal story behind a code change while the work is happening: intent, decisions, rejected approaches, surprises, and lessons. Each problem-solving arc lives under docs/work-backstory and links to related commits in both directions. A plugin can add a one-line compaction checkpoint, but continuous capture remains the main protection against context loss.

launch-doctor-kr scans a Korean Android app before Google Play submission, with checks for missing Maps keys, outdated targetSdk, release debuggable flags, Capacitor Hangul input issues, location-service reporting, AI-content labeling, and privacy-policy gaps. Findings include a file location, a suggested fix, and a supporting link. It is aimed at small teams using Capacitor, React Native, or Flutter.

Provide a Four Pillars chart, gender, and luck-cycle details to convene several competing BaZi schools. They review the chart in rounds, challenge one another, and make predictions with probabilities, time windows, and explicit failure conditions before you reveal outcomes for scoring. The Chinese-language workflow is intended as a reflective reasoning exercise, not health, financial, or legal advice.