Recently updated open-source projects

Interview Mapper builds qualitative analysis from atomic, quoted evidence instead of summaries of summaries. It verifies that quotes are verbatim and checks whether each quote actually supports the attached claim, then assembles insight cards, personas, journey maps, opportunity prioritization, or decision memos. English and Russian packages include sixteen analytical lenses, five output types, and reliability checks.

Explain Architecture scans the important parts of a codebase, such as manifests, entry points, migrations, and routes, then produces readable Mermaid diagrams for modules, data flow, ERD, request lifecycle, and deployment. It also writes a plain-English ARCHITECTURE.md, an AI-CONTEXT.md brief, and editable diagram sources, with PNG and PDF renderings for people to review.

Peanut Illustrations turns a one-sentence idea into a metaphor, composition, and recurring deadpan peanut character, then generates one white-background 16:9 concept image and adds the watermark locally. It favors one core structure, generous whitespace, and very little text rather than infographics or slide graphics. Image labels can still contain errors and should be checked before sharing.

WebGL Animation Skills helps an AI coding agent build interactive 3D scenes, generative backgrounds, and particle effects for the web. Its GLSL, Three.js, and particle-system skills cover shaders, GLTF motion, scroll-linked scenes, and GPU resource disposal, with practical paths for common WebGL work. Target projects still need device-specific performance and memory testing.

The Council sends project context and a focused prompt to Codex and Gemini in parallel, then has Claude synthesize their agreement, disagreements, and a final recommendation. It supports code review, architecture decisions, debugging, and general second opinions, with single-advisor fallback when one CLI is unavailable. Versions 1.2.0 and 1.2.1 did not reliably enforce Gemini read-only access, so version 1.3.0 or later should be used.

video-reader drives ffmpeg and ffprobe to extract frames, metadata, spectrograms, and waveforms from local media, with optional Whisper transcription and yt-dlp URL downloads. It exposes both standalone Python scripts and a Claude Code skill, while the core features require only ffmpeg. Frames are downscaled and capped by default so visual inspection does not overwhelm the context.

hekouwang-stock-data-reader-skill accepts an A-share code, pulls public market and company data through AkShare, calculates a fixed set of indicators, and produces publishable chart cards plus long-form and headline versions. Its built-in boundary is deliberately non-predictive: it does not recommend stocks, forecast prices, or give trading points, and adds a risk notice to the output. The data source needs no API key, but the result remains an informational retrospective rather than investment advice.

Enterprise Architecture Team helps an organization establish and run an enterprise architecture function, from its charter and target-state vision to principles, standards, stakeholder engagement, maturity assessment, portfolio rationalization, and decision records. It is aligned with TOGAF while remaining vendor-neutral and can route a request to one artifact or a complete operating blueprint. A matching GitHub Copilot prompt file keeps the workflow available outside Claude.

video-digest processes local videos into Markdown digests containing spoken transcripts, on-screen text, URLs, handles, and timeline contact sheets. ffmpeg is the only hard requirement; faster-whisper and OCR engines are optional and degrade gracefully when absent. Independent stages cache their results and can resume, and the tool never uploads media or needs an account.

Goalkeeper drives a repository task through repeated build and independent verification rounds, declaring completion only when machine-checkable acceptance criteria pass. It protects previously green work by reverting failed or gamed rounds, and halts on stuck items, no progress, oscillation, or configured budgets while writing a durable escalation report. It requires Claude Code's dynamic workflows, a Git repository, and acceptance checks that can produce objective pass or fail results.

PCB STEP Slim removes electrical and cosmetic detail that makes PCB assemblies heavy while keeping the board outline and mechanically relevant components, including height-critical parts. Its reference-graph approach preserves a valid, closed STEP structure, original geometry, and component colors instead of re-tessellating the model. The result is intended for faster mechanical review in SolidWorks or other CAD tools.

Hemingway Skill reviews marketing copy, technical documents, blog posts, specifications, and other writing for clarity and economy. It flags long or complex sentences, passive voice, weakening adverbs, jargon, hedging, and empty constructions, then explains the issues and proposes rewrites. Outputs can include readability statistics, a fully revised version, and takeaways, with modes for quick review, focused checks, shorter copy, or specific content types.

Immos is a local immigration-preparation assistant that helps organize a person’s situation, compare visa pathways, inspect missing documents, plan deadlines from current official information, and draft application letters. It keeps the case files on the user’s machine and does not submit forms or act as a lawyer. Legal judgment, eligibility decisions, and complex histories remain with the user and an immigration attorney.

DecentPaste lets computers and phones share clipboard content directly over the same Wi-Fi or LAN without accounts, subscriptions, or a central cloud service. Devices discover one another automatically, then pair with a six-digit PIN before exchanging end-to-end encrypted data. Prebuilt releases cover Windows, macOS, Linux, and Android; iOS is still listed as forthcoming, and mobile devices must keep the app open while pairing.

Translate Ru En converts Russian prose, Markdown, fiction, technical specifications, chats, or directory trees into natural English. It preserves headings, code blocks, links, identifiers, tables, numbers, formatting, and the author’s voice, translating meaning rather than word-for-word calques. Source instructions are treated as text, not executed, and code or technical names stay unchanged.

Nex turns agent conversations and connected business sources into a shared knowledge graph, so facts recorded in one tool can be recalled from another without repeated explanations. Its setup command detects supported coding platforms and installs hooks, MCP integration, and slash commands, while a hosted MCP endpoint supports other clients. Using the service requires a Nex account and any desired data integrations.

Adversarial Design Review convenes specialist agents to challenge a difficult design before code is written. They look for concrete failure scenarios in drivers, on-disk or wire formats, concurrency, protocols, and the CI checks meant to prove correctness; you reconcile the findings and apply fixes while they are still cheap. It is aimed at work that is hard to debug or reverse after implementation.

This portable pack contains 25 Agent Skills for recurring engineering work such as task breakdown, test-driven development, debugging, security review, migrations, documentation, pull requests, and deployment. Each skill defines completion checks and explicit points for human judgment. The same directory layout can be used with Claude Code, Codex, Cursor, and other compatible assistants.