Recently updated open-source projects

craftcode gives an AI coding agent seven defaults: mirror neighboring code, write intent comments, fail open only at safe edges, make money and auth fail closed, preserve idempotency, prefer native code for tiny needs, and keep a defensive boundary around a clean core. It runs a final checklist and explains when to bend the rules for prototypes, house style, or readability.

Run the skill from Claude Code or Codex to wait for Gemini Code Assist, fetch unresolved actionable review threads, apply clear fixes, verify them with repository checks, push, and request another review. It distinguishes resolved, outdated, and reply-addressed threads and stops at a configurable rereview cap. The workflow requires Python 3.10+, an authenticated gh CLI, and a pull request configured with Gemini Code Assist as reviewer.

Use this methodology to turn an AI initiative into a measurable, stage-gated plan that can stop before expensive product work begins. It decomposes capabilities, ranks risks, isolates one major variable per stage, defines human validation and acceptance metrics, and sets explicit pass, fail, and stop-loss conditions. The central rule is to prove the core AI capability before building the surrounding platform, interface, database, or automation.

Use this Claude Code skill to formalize requirements, check compliance and compatibility, research current price bands, calculate total ownership cost, and compare vendor risk before buying. It scales the decision method to the importance of the purchase, supports both consumer and business procurement, and produces a cited memo with post-purchase steps for inspection, warranty, maintenance, and eventual resale or retirement.

epiage-skill computes 25 epigenetic aging clocks from a CpG beta-value file for human whole-blood samples, including GrimAge, Horvath, Hannum, PhenoAge, DunedinPACE, and DNAmTL. It uses only pandas and numpy offline, can fill missing CpGs with a vendored methyLImp reference, and reports acceleration plus per-clock reliability flags. The input must match a supported blood methylation export; other tissues are outside scope.

Ask for three, five, or ten Indian stock ideas and choose an investment horizon. NSELens screens the Nifty 100, verifies prices and analyst targets from named sources, and combines fundamentals, technicals, sentiment, catalysts, and risk/reward. Shorter horizons receive a GTT-style trigger, target, and stop-loss setup; longer horizons receive entry zones. The output is research, not financial advice.

cinematic-ai-film teaches an end-to-end workflow for finishing an AI promo film, from locked beats and image-to-video shots to one consistent voice, real product UI capture, sound design, ffmpeg assembly, and measured overlays. It focuses on the last-mile failures that ruin a deliverable—continuity drift, mispronunciations, mask alignment, and runaway renders—and records costs and hard-won fixes from a real launch film.

Add a focused web-animation toolkit to AI coding agents such as Claude Code, Cursor, or Codex. The pack covers GSAP timelines, SVG and Lottie motion, micro-interactions, page transitions, glass effects, and performance practices for 60/120fps rendering. It also includes reduced-motion and accessibility guidance, making it useful when animation must feel polished without causing layout jank or ignoring user preferences.

uxlaws gives a design agent a catalog of 30 UX principles, each paired with its mechanism, application guidance, pitfalls, and a real example. A compact symptom-to-principle index routes the relevant entries and loads the deeper catalog only when needed. Use it as a focused UX pass during interface design or review, not as a generic style guide.

Pine Script v6 skill helps Claude Code write, debug, review, and migrate TradingView indicator code without inventing function signatures or confusing v5 with v6. A lookup script queries the offline reference for built-ins, typed arguments, return values, examples, and common v6 gotchas, while migration notes cover frequent version changes. The skill still treats the live compiler as the final check and requires Node.js 18+; its reference data is fetched on demand from a pinned upstream revision and integrity-checked.

keepp-skill lets an agent manage a Keepp page through the Keepp Agent API. It can create, list, update, or delete product cards; change layout, about text, theme, hero actions, and card defaults; and choose between an external link, a detail view, or lead capture. A Keepp Pro API key is required.

founder-decision reframes the decision, distinguishes reversible from one-way doors, selects only the mental models that fit, and runs a pre-mortem against known startup failure patterns. It ends with a clear recommendation, the assumption carrying it, and the signal that would change the answer. An optional gitignored context file can anchor the analysis to your market, stage, model, and constraints.

image-studio turns a plain-language design request into HTML and CSS, then renders it as a sized PNG or JPEG with headless Chromium. It supports ads, social cards, thumbnails, and other static formats, while its reverse-template workflow can convert an existing ad screenshot or URL into a parameterized template for versioning and batch variants. It requires Node.js and Playwright's Chromium download.

brand-kit extracts a client’s visual identity into a validated DESIGN.md and keeps colors, typography, spacing, and component tokens in a central registry. Scripts flag unknown values instead of guessing, lint contrast against WCAG AA, detect drift, and sync approved tokens into projects such as image or landing-page work. Unverified items require confirmation before the file becomes authoritative.

zip-code-research takes a service radius, combines it with US Census ACS data and ZIP boundary shapes, and produces Include, Test, and Exclude targeting lists plus an interactive map. It is aimed at home-services financing teams that want to choose markets more deliberately and reduce credit denials. The scoring rationale can be adapted by vertical, while the underlying script uses only Python’s standard library.

Answer guided questions about your name, positioning, proof, products, and profiles, then use the skill to create a crawlable entity site with structured data, consistent bios, backlinks, and indexing files. It also walks you through IndexNow, Search Console, and checks across search-connected assistants. The process can improve discoverability after indexing, but it cannot guarantee immediate citations or change an offline model's memory.

solana-protocol-architect-skill keeps coding agents in the implementation role while it designs the protocol first. It examines oracle, collateral, liquidation, authority, and economic-attack choices, then produces a threat model, parameter review, runnable invariants, and tests for an implementation agent. It runs from a protocol description without API keys or blockchain access.

Not So Fast checks whether an agentic decision was verified with enough independence for its difficulty and material impact. A self-only critic cannot approve a high-materiality, hard-correctness call; the guard instead requires an independent check or escalates when none is available. It ships as a Claude Code plugin, a SKILL.md folder, and a dependency-free Python guard with deterministic verdicts. The project presents this as a reference contract, not an empirically validated performance mechanism.