Recently updated open-source projects

Timeback notices where work sessions consume time, offers an optional five-minute drill between tasks, and saves the result as a named playbook that can be run again. Practice starts with fake data, and success means the new habit sticks rather than merely receiving a positive response. The early release ships with a cold-email example.

electron-util packages common helpers for Electron applications into separate shared, main-process, and pure Node.js entry points. Developers can use it for environment and process detection, development-mode checks, and ASAR path handling without duplicating small utilities. Main-process features still require an IPC layer when exposed to a renderer, and the current package requires Electron 37 or later.

Dope reads a person’s communication tendencies through four bird profiles—Peacock, Dove, Eagle, and Owl—and uses the selected profile to shape everyday messages, sales conversations, and relationship advice. The skill keeps the model, reading guide, and three playbooks in separate reference files, so an agent can load the relevant context for a situation. It is best used as a conversational hypothesis about how someone may receive a message, not as a permanent label.

Next.js App Router Audit scans an App Router codebase with a zero-dependency Node tool for Server Component boundary mistakes, request waterfalls, client-side anti-patterns, and Core Web Vitals risks. It reports issues such as lost metadata, legacy head or image usage, needless client components, effect-based fetching, raw images, unsafe HTML, and missing streaming fallbacks, with severity-ranked fixes. A companion checklist covers judgment calls the scanner cannot prove, and JSON output can fail CI on errors.

Log meals from photos or text, record exercises with sets, reps, and weight, and calculate BMR, TDEE, daily intake, expenditure, and calorie targets. It can also generate weekly training and diet plans for cutting, bulking, or maintenance; records stay in local JSON, while AI features require an OpenRouter API key.

Codex Session Bridge keeps Codex Desktop conversations visible when switching between a personal ChatGPT account and an API proxy. It backs up and updates local JSONL sessions and the SQLite index, with dry-run, verification, diagnosis, and rollback commands. Codex Desktop handles the API key through its prompt and macOS Keychain; Codex must be restarted after a bridge.

Give Podcast Workflow a topic and it researches angles, writes a conversational script, pauses for your approval, and synthesizes the finished audio through Xiaomi MiMo TTS. You can use built-in Chinese voices or clone a voice from a 30–60 second sample, with paragraph-level emotion cues. It requires Claude Code and a MiMo API key.

ai-landing-skill teaches compatible coding agents how to compose performative-ui’s React components into AI-startup landing pages. It includes recipes for hero, social proof, pricing, and waitlist sections, three worked examples, and rules such as using one hero centerpiece, keeping metrics plausible, and checking mobile layouts. The visual language is intentionally AI-startup oriented, so the skill defers for regulated industries, generic non-tech sites, app interfaces, or projects with an existing design system.

Arcana installs a large library of coding skills and curates which ones enter an agent's active context based on the project, past usage, and a configurable token budget. Its CLI also compresses noisy command output, stores cross-session memories and snapshots, and manages selected MCP servers. It is aimed at developers working across tools such as Claude Code, Codex, Cursor, Gemini, and Aider who want useful context without loading every skill at once.

Use it when a repository's open issues no longer make the next step obvious. The skill reads issues through the authenticated gh CLI, classifies what is ready, blocked, or underspecified, computes a dependency-safe order, and renders a self-contained HTML map showing which issue chains unlock each feature. It requires Python 3, gh, and a modern browser.

Point codemap at a repository to build an interactive map of functional modules, dependencies, and file-level findings, with a technical-debt score for each module. Incremental runs re-audit only changed areas, while the fix workflow locks a test baseline and requires an independent acceptance check before re-scoring. Module boundaries and scores are model judgments, so the output is intended as a reviewable maintenance starting point rather than deterministic static analysis.

Turn a small image or merged PSD composite into an exact-size, DPI-tagged TIFF using tiled Real-ESRGAN passes and final resampling. The CLI can target inches or pixels, handle aspect-ratio choices, and emit a JPEG proof; a browser version runs locally with a lighter model. CUDA or Apple MPS is recommended for large canvases, and the process cannot recover true native detail.

Advertorial Agent takes a conversational brief through intake, copy, design, images, assembly, and a Framer CMS push, pausing for approval at four checkpoints. It retrieves niche-matched examples for headlines, voice, structure, calls to action, and image prompts, then produces a parent row plus ordered section rows. The project is a single-user Claude Code setup that requires configured language, image, vector, and Framer services.

Map a codebase to produce a human-readable CODEDNA.md profile covering naming, comments, structure, control flow, error handling, and recurring idioms. Coding agents can then use that profile while writing new code or checking a file or diff for stylistic tells. The analysis focuses on choices formatters do not control, rather than duplicating rules already enforced by tools such as Prettier or Black.

Claude Nginx Skill covers nginx authoring, review, hardening, and debugging for virtual hosts, reverse proxies, WebSockets, TLS, HTTP/2 and HTTP/3, SPAs, object storage, load balancing, caching, and rate limits. It diagnoses from access logs before proposing changes and includes guidance for mobile MTU failures, VPN asymmetry, and version-specific syntax. Its two shell scripts are read-only; changes should be checked with nginx -t and a reload.

P2H2P converts an old deck, rough idea, whiteboard photo, or screenshot into editable HTML and PPTX. It orchestrates template selection, parallel HTML building, visual checks, conversion, and final QA, pausing at outline, template, and HTML-preview checkpoints. It is an orchestration layer rather than a standalone renderer and needs the companion HTML-template and PPTX skills plus their runtime tools.

Study Abroad Planner asks for age, language scores, education, budget, destination, and constraints, then produces a school shortlist, EOI or visa pathway, cost and cash-flow view, risk register, and near-term actions. Australia has the deepest coverage; the UK, New Zealand, and Canada are shorter references. It does not file applications or guarantee admission or visas, and it is not a substitute for a licensed migration adviser.

stop-slop-fr reviews French academic reports, theses, dissertations, and research writing for repetitive connectors, empty formulas, mechanical structures, and unsupported source language. It adapts tone to the assignment, asks for concrete author-and-year citations, and uses a qualitative 50-point score to flag text for revision. It does not replace grammar tools or human supervision, does not bypass commercial detectors, and may need discipline-specific judgment.