Recently updated open-source projects

threat-model-studio reconstructs data flows and trust boundaries, enumerates STRIDE threats, and applies LINDDUN to personal-data flows. It scores risks through a deterministic OWASP rubric, maps them to mitigations or explicit accept or transfer decisions, tracks regressions in a versioned register, and exports Threat Dragon JSON plus Markdown.

council-of-neurodivergents sends one question to four advisors with distinct neurodivergent cognitive lenses, lets them cross-review one another anonymously, and asks a Chairman to synthesize a verdict and next steps. It is a compact Claude skill for challenging assumptions through multiple perspectives rather than receiving a single first-pass answer.

X Account Teardown reconstructs an X account’s growth arc from exported public posts, highlighting inflection points, content pillars, cadence, reply behavior, hook patterns, and a growth-curve proxy. It can pull cookies from a logged-in Chrome profile and work around the usual tweet cap with date windows. Follower curves are not public, deleted or followers-only posts are unavailable, and the project is for research.

Run `/savenow` in an OpenClaw Telegram topic to preserve resolved fixes, conventions, decisions, and system mappings without keeping the surrounding conversational noise. The skill compares candidates with the day's memory, shows an exact Markdown preview, and writes only after an Apply confirmation unless auto mode is chosen. It is intentionally limited to OpenClaw with Telegram, and each workspace can hold only one pending preview at a time.

embedded-code-logic helps structure bare-metal MCU and FreeRTOS firmware around state machines, ISR-to-task cooperation, driver ownership, timeouts, and watchdogs. It reviews races, deadlocks, priority inversion, and shared-resource protection, while deliberately leaving register settings and datasheet lookups to the chip documentation.

forex-forecasting-skill answers plain-language questions about currency pairs or gold with live MetaTrader 5 data. It combines D1, H4, and H1 trend, momentum, support and resistance, volatility, and macro context into a bullish, bearish, or neutral bias with invalidation levels. It can render a shareable candlestick card and add a trade setup only when requested; it never places or changes orders.

weekly-news-seo gathers a niche’s weekly stories, curates a small shortlist, and drafts original articles for classic search and AI answer surfaces. Each article adds context beyond its sources and can include direct-answer blocks, question headings, sources, structured data, and internal links. You approve or reject every draft before it is written into the site, committed, and deployed through Vercel; the workflow is intentionally low-volume rather than an auto-publisher.

Tokendiet converts PDFs, HTML pages, URLs, or images into lean Markdown and reports estimated token and cost savings. It includes OCR support, reproducible benchmarks, and warnings about visual loss, but charts, diagrams, complex layouts, and other visually important content are better read natively rather than reduced to text.

ATS Resume Reviewer compares a resume with a complete job description through a 13-module ATS and recruiter-style report. It covers parsing, compatibility, keyword gaps, experience, formatting, risks, and a prioritized improvement roadmap, while refusing to add skills, metrics, certifications, or achievements that the candidate did not provide.

claude-vastai-skill provides a Claude Code and terminal control plane for searching, renting, watching, using, syncing, and destroying Vast.ai GPU instances. Each rental can carry an auto-destroy deadline guarded by a watchdog, with cost and orphan checks; the CLI wraps the official Vast.ai tool and still requires an API key.

kickoff interviews the user to turn a one-line idea into a PRD and a cluster-based plan with measurable success criteria. The user approves before coding, Codex signs off each cluster as it lands, and a final regression pass plus an over-engineering audit closes the work. An Obsidian folder preserves detailed state across sessions. The skill expects related planning, frontend, audit, and authenticated Codex tools to be installed.

Describe the capability you need in Chinese and the skill searches a catalog of more than 88,000 agent skills, applies seven quality checks, and suggests options that can be installed directly. Its recommendations emphasize Chinese workflows such as Xiaohongshu, WeChat official accounts, and Feishu, and the curated list can update remotely without reinstalling the skill.

Use Rosetta to combine local conversation histories from 18 coding agents with git and project documentation, producing a cited snapshot of what is actually true now. It can also create durable architecture, product, and business decision records with provenance and status changes. The core Python CLI runs locally with no third-party dependencies and reads transcript stores and product code by default, but its normalized cache may contain secrets and should remain git-ignored.

Use this Claude Code skill to scan a Plex library, find promising titles that are not yet owned, and explain each recommendation through movies already on the shelf. It publishes recommendation cards and an interactive proximity map, preserves earlier rounds, and updates the taste profile when suggested titles are later added. Plex is required; Tautulli adds per-user views, while Overseerr or Jellyseerr adds availability and request actions.

single-source-content-factory turns one transcript, draft, talk, or document into a canonical asset with addressable content atoms. It adapts those atoms for blogs, newsletters, LinkedIn, X, video scripts, and email, checks voice and claim consistency, emits a traceability matrix, and identifies only the outputs affected by source changes; it plans cadence but does not post or schedule.

eval-harness-architect starts from the decision an evaluation must support, then builds versioned datasets, holdouts, dimension-specific scorers, and a calibrated judge. It measures significance, wires regression gates and PR scorecards into CI, and adds drift and production-feedback loops without fabricating labels or benchmark numbers.

talk-and-keynote-architect builds the spoken narrative before slides: one through-line, a timed beat sheet, an ear-friendly script, hostile Q&A preparation, and iterative rehearsal cuts. It measures the real runtime and produces a day-of kit, while routing the actual deck to a separate design layer.

monetization-architect carries pricing work from value-metric discovery and willingness-to-pay studies through tiers, revenue scenarios, and a safe live migration. It defines grandfathering, cohorts, churn guardrails, rollback triggers, and modeled-versus-actual reconciliation from user-supplied data, rather than reading a billing system.