Shape agent capabilities into your workflow.

Describe a project in plain language and `/scaffold` routes the request to a stack-specific wizard, shows the complete plan, and waits for confirmation before generating code, configuration, and project guidance. `/bootstrap` surfaces relevant community skills, while `/sync` exports project memory to Git for reuse across machines and several coding tools. The collection includes more than 30 skills, although many topic-specific reference skills are still clearly marked stubs.

Start with a few questions about priorities, voice, audiences, and campaigns, then crawl the company site and research competitors and public reputation signals. The skill maps the findings to HubSpot's Business, Customers, and Brand Kit fields, respecting form constraints, and produces a tabbed HTML paste sheet with individually copyable values. It can also export product research, ICPs, and personas for future updates.

email-design guides Claude from brand context and a few high-value questions to one of six committed email archetypes, a structured brief, real imagery, and a sending-tool-ready layout. It avoids generic palettes and filler images, while specifying asymmetric spacing and a real footer with unsubscribe details. Nitrosend is the first-class integration, but the brief can be translated for other MCP-enabled ESPs; the skill itself does not send mail.

Ask what is new today and the skill gathers signals from Product Hunt, GitHub Trending, and Hacker News, then weights them against recent reports and project context. It produces a 60-second summary, three to five deeper reads, priority labels, and cross-source patterns in Markdown, with optional PDF output through gstack. Claude Code WebFetch access is required, and a run typically takes a few minutes.

yss-overtime-submission reads attendance and effort data from ZenTao, applies mainland China workday, holiday, and makeup-workday rules, then compares existing Kingdee s-HR submissions before filling the form with Playwright. It supports dry runs and splits days longer than eight hours into multiple entries, but real submission requires an explicit --confirm after a person logs in. The default selectors target ZenTao plus Kingdee s-HR Cloud.

This Claude Code skill improves a repository's public presentation by rebuilding the README from verified project facts, refining the GitHub description, and assigning a focused set of topics. When explicitly requested, it can initialize and push a new repository; it can also add a LICENSE file when the README declares a license but the file is missing. It does not change application code, and applying GitHub metadata requires an authenticated gh CLI.

Claude Harness scaffolds a repository-level working framework for AI-agent work. It creates concise rules, active and completed execution plans, lessons, and a technical-debt tracker, then uses failing tests to lock behavior before implementation. Review, phase-finish, and lesson-promotion steps keep the context current over longer projects, making it useful for new codebases, refactors, and durable maintenance.

agency-os turns Notion into a visual control board where ideas are discussed, scoped into dependent tasks, approved by an operator, and then assigned to agents that return result links to the same rows. Nothing runs autonomously: only explicitly marked work is dispatched, making it suitable for founders and teams that want visible status and human control; setup requires a Notion integration and an MCP-capable host or Claude Code plugin.

Design Review evaluates a page by looking at rendered screenshots rather than guessing from source code. Through Playwright MCP it captures desktop, tablet, and mobile views, checks typography, contrast, spacing, hierarchy, color, alignment, and responsive behavior, then proposes ready-to-apply CSS fixes. It can rerun the audit after changes, but requires a connected Playwright MCP server.

academic-topic-research-agent turns a one-sentence research idea into a preliminary package for venues such as CHI, CSCW, LAK, and AIED. Quick, standard, and deep modes can produce literature metadata, feasibility verdicts, title and abstract options, research questions, method boundaries, evaluation plans, risks, next steps, venue positioning, and reviewer-attack predictions. A concept stress test adds multi-angle critique and revision; it is not a paper-writing service or a substitute for scholarly judgment.

ai-video takes a request from first-frame generation and prompt writing through rendering, clip review, voiceover, assembly, and upscaling. It routes a shared request format to six supported video models and records lessons from each quality-control pass. You need API keys only for the services you call, plus ffmpeg and ffprobe locally.

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.

This skill audits a page for visibility in Google AI Overviews, ChatGPT Search, Perplexity, Gemini, and Copilot. It checks foundational SEO, debunked GEO tactics, extractability, authority, and citation readiness, then scores three pillars and proposes prioritized fixes; it reports what cannot be assessed without page access or performance data.

Run a two-stage cleanup on German text: deterministic Python scripts remove or normalize hidden Unicode characters, then a linter and language model flag formulaic cadence, puffery, shallow analysis, and other AI-like patterns. The skill returns structured findings and can rewrite the text without inventing missing facts. It is tuned for German and explicitly does not promise detector evasion or removal of statistical watermarks such as SynthID.

IRIS gives Claude Code a disciplined network-operations workflow for discovering device state, checking it against a source of truth, preparing a change, verifying the result, and documenting the ticket. Engineers retain the decision, while connected MCP tools, network documents, and discovery scripts supply context. Read-only automation can assist investigation; state-changing actions require explicit approval.

cuecards assigns or auto-detects a profile for each project directory, then launches Claude Code or Codex with only the relevant skills, MCP servers, plugins, persona, and quality gates. The same profile can also be materialized into native configuration files for Cursor, Cline, and other supported tools. It reduces irrelevant always-on context and global configuration sprawl, requires Node 20 or newer, and acts as a configuration and launch layer rather than an agent replacement.

Enter a date to collect public ETF fund-flow data and receive a structured view of broad-market direction, sector rotation, cross-border sentiment, and an overall buy, sell, or hold signal. It is designed for investors and quantitative researchers using flows to confirm trends rather than predict them, and its output is research material rather than investment advice.

This skill turns clinical .xlsx or .csv data into a standalone Korean HTML exploratory-analysis dashboard with dataset summaries, missingness, distributions, outlier candidates, optional Table 1, correlations, and VIF. It embeds assets locally, masks identifier columns, and supports print or PDF output, but does not perform inferential analysis or handle images, waveforms, or free text.