Recently updated open-source projects

ralph-loop-setup wraps the Ralph loop for Claude Code or Codex: write a PRD, split it into atomic stories, and let a fresh agent implement one unfinished story at a time with checks and commits. Progress persists in prd.json, progress.txt, and Git history. A branch is not done until both the L1 PRD-focused review and an independent L2 model review are green, or every finding is explicitly closed.

Scan Solana code for return-data spoofing, arbitrary program invocation, stale account state after CPI, and unsafe PDA signing. The skill routes findings to Anchor or native Pinocchio guidance, explains the exploit path, and proposes concrete remediations through an audit command and dedicated agent. Runnable LiteSVM proof-of-concept suites demonstrate vulnerable, defended, and positive-control cases, with especially detailed coverage of spoofed CPI return data.

Describe an image or provide a weak existing prompt, and this Claude skill will ask up to three questions only when a critical choice is missing. It then builds one prompt through a ten-part structure covering composition, subject, lighting, palette, camera or typography, finish, and references, with adapters for GPT Image 2 and Nano Banana. It writes prompt text only and does not call an image-generation API.

legal-launch-review checks Chile-focused AI, SaaS, RAG, API, and web projects for personal-data handling, privacy notices, terms, basic security, AI risks, e-commerce duties, and licenses. It reviews the actual code and returns a simple red-yellow-green launch view with located findings and suggested fixes. The scope is advisory and limited to Chile, so laws, citations, and launch decisions must be verified with official sources or counsel.

Revolt Desktop contains the Electron client for using and developing the Revolt chat service on a desktop. Contributors can install dependencies, build the application bundle, run it locally, and produce platform-specific packages for Linux, macOS, and Windows. The repository is useful for maintaining the existing client, testing desktop behavior, or submitting fixes. It also carries an explicit deprecation notice: a new Tauri-based desktop app is intended to replace this project, which matters for anyone considering long-term development on the current codebase.

assay interviews the user until scope, repositories, contracts, and testable criteria are explicit, then runs an autonomous plan, research, implementation, convergence, polish, and ship-gate workflow. Machine checks decide whether correctness and completeness are sufficient, while the human intervenes at intake and before outward actions such as commits, pull requests, or ticket writes. The skill adapts to backend, frontend, mobile, CLI, library, and infrastructure projects across one or several repositories.

paper-ingest resolves a paper URL, title, or DOI into metadata, extracts full text from the PDF, classifies the work, and renders a detailed Obsidian note. It can push references into Zotero, detect duplicates, and collect PDFs for later attachment through WebDAV. The skill expects a local config containing service credentials, so it fits researchers who already use Zotero and Obsidian.

Project Foundations adds language-specific setup guidance to Codex and Claude project work without replacing the existing superpowers workflow. Its references cover project structure, naming and style, tooling, starter configuration, TDD, CI checks, quality gates, and common review points across many languages. An installer registers the plugin with Codex and places the marketplace entry, while the same package can be installed for Claude. It is useful when new projects need consistent starting conventions.

universal-audit-skill classifies almost any work product, selects relevant verification domains, and produces a numbered audit record with severity-rated findings and remediation links. It supports code, applications, APIs, architecture, infrastructure, reports, spreadsheets, and media across Claude, Codex, Cursor, Gemini, and Copilot. Constructive and adversarial passes can be combined, while external standards and facts still need current validation for the project.

stock-research-skill offers Chinese and English variants for multi-source equity research in Cursor and Claude Code. The playbooks combine Schwab market signals, SEC filings, Polygon news, and ClickHouse data for shakeout-with-news and insider-alert analyses, with health checks and version gates before use. The repository is documentation-only; connected MCP servers own the actual API traffic.

rename-papers scans a folder, reads embedded metadata and opening-page text, then verifies canonical author, year, title, and DOI through CrossRef before proposing a filename. It silently skips non-scholarly documents, sends ambiguous papers to a needs-review group with reasons, and preserves tags such as [ANNOTATED] or [v2]. The tool is beta and applies changes only after a dry run and confirmation.

Multi-Agent Workflow Kit creates a separate Git worktree, branch, and directory for each coding agent, then places them in split panes within one tmux session for supervision and messaging. Agents are configured in .agents/agents.yaml and managed with maw commands for provisioning, syncing, navigation, and communication. The project is still an early proof of concept and requires Git 2.5+, tmux 3.2+, yq, and uvx.

expense-reimbursement-skill reads receipt folders and reimbursement templates, backs up originals, classifies and renames materials, de-duplicates one expense into one ledger row, and fills a copy of the template chronologically. It reconciles totals and places unclear or unrelated items on an open-questions list, while keeping original files untouched and supporting invoices, PDFs, OFD, images, and spreadsheets.

Lead Magnet Architect guides a lead-magnet build instead of generating a one-shot asset. It tests whether a stranger would pay $50 for the standalone resource and whether the small win exposes the larger paid problem, then routes the idea to a format checklist, reveals steps incrementally, and runs a final quality pass.

3d-brain-animations turns a sentence such as “rotating rat hippocampus” into a PNG figure or MP4 animation. It selects a BrainGlobe atlas, finds the named regions, builds a translucent brain scene with glow and lighting, and renders it headlessly in Blender so the agent does not need to write 3D code. More than 200 atlases support multiple species, making it useful for papers, teaching, and talks. Blender and atlas data remain required runtime dependencies.

Import chat records to score six dimensions including initiative, response quality, uniqueness, emotional disclosure, behavioral signals, and future-oriented language. The app visualizes relationship trends, highlights notable lines, and flags possible over-interpretation or one-sided interest. Records are parsed locally, but a text summary is sent to the DeepSeek API for analysis; the result is intended as a personal reference.

teams-video-dl-skill extracts a Teams recording and its speaker-labeled transcript from a signed-in Chromium browser session, then uses ffmpeg to mux the video and audio tracks. Scripts run through uv, so no manual Python dependency setup is needed. Processing stays local; temporary HAR and curl files can contain session cookies and tokens and are removed after a successful run, making the local machine boundary important.

mpp-to-xlsx converts a Microsoft Project .mpp schedule into a branded six-sheet .xlsx workbook. It carries over the work-breakdown hierarchy, tasks, team assignments, effort, amounts, notes, and a daily resource plan, then adds a resource report crossing people, tasks, and months. Conversion requires Java 8+ plus Python 3.9 and the listed Python packages.