summit-claude-skills

Reusable 6-role marketing prompt framework for Summit. Treats Claude as a six-role functional marketing team (Market Researcher, Brand Strategist, Content Strategist, Copywriter, Growth Analyst, Repurposing Engine) wired by a System Flow and structured via the Operator Prompting formula (Role/Context/Objective/Task/Output). Activate when user says "marketing team", "run marketing flow", "act as [role-name]", "research the audience", "find positioning", "build content pillars", "write hooks/scripts", "analyze post performance", "repurpose this content", or asks for any structured marketing artifact (pain-desire map, positioning statement, content pillars, hooks, repurposed assets).

Claude-Code-only video editing pipeline. Turns raw footage + a one-line prompt into a fully edited short-form video with motion graphics, subtitles, and music — without Premiere, DaVinci, or CapCut. Stack: HeyGen HyperFrames (HTML→video render engine, headless Chromium + paused GSAP), browser-use/video-use Claude skill, and nateherkai/hyperframes-student-kit reference workbench. Reverse-engineered from @cooper.simson IG reel (DXiyhqIjn_n). Activate when user says "edit this video", "cut the umms", "make a reel", "add motion graphics", "render the final cut", or hands over raw footage + a brief.

Extract the full design DNA of any website — layout, color scheme, typography, animations, and characteristic elements — into a fixed-schema Markdown design spec using Claude in Chrome, captured across desktop and mobile views. Use this skill whenever the user provides a URL and wants its design extracted, cataloged, analyzed, or used as a reference for a new build. Activates on "extract this site's design", "website-design-extractor", "catalog this site", "I want a site like {URL}", "pull the design from {URL}", or any request to document how a website looks and behaves.

Pre-decision Socratic checklist — 7 prompts that force the user (and Claude) to challenge a position before committing. Anti-vending-machine habit: bring the thinking, let the prompts attack the premises. Activates when the user says "socratic check", "pre-decision", "before I commit", "stress my reasoning", or presents a decision and asks to pressure-test their own logic (not the idea itself — use devils-advocate for idea attacks).

Pack the active tasking of the current Claude.ai chat into a clipboard-ready, token-efficient summary for handoff to a fresh Claude.ai session. Captures only what's needed to continue the active work, know the next action, and know when the work is complete — nothing archival, nothing conversational. Activates when the user says "context summary", "context-summary", "pack the context", "pack this context", "fresh context handoff", "context handoff", "summarize this session", "wrap this for a new chat", or any variant. Output is a single fenced markdown block at the end of the response so it can be tap-and-hold copied on mobile.

Designs a system from concept to implementation-ready blueprint across four system classes — user interface, code architecture, business process, and automation pipeline. Runs a structured five-phase workflow: scope, identify parts, define interfaces and flow, specify failure and evolution paths, then hand off to a build skill. Use when the user says "design a system for...", "design this system", "how should this system work", "sketch a system for X", "what should this system look like", or when the user is about to build something non-trivial and has described the outcome but not the parts. Sibling to `systemize` (promotes an existing action into a system component) and `architect-plan-for-dispatch` (produces an execution plan once a design exists). This is the thinking step before those.

Capture, structure, and organize ideas from any source (URLs, screenshots, text, voice transcripts, calendar inbox events, iOS Shortcut payloads) into the ryankolean/mindspace GitHub repo. Classifies along two axes (project + idea-type) with emergent tags, auto-schedules High-priority items via Google Calendar reminder-hub conventions, and self-improves by appending to learnings.md after every session. Activates on "brain dump", "capture this", "add to mindspace", "process my inbox", "mindspace review", "what did I capture about X", or when the user shares a URL/screenshot/idea fragment without explicit instructions.

Package the relevant content from the current Claude.ai chat into a clipboard-ready, self-contained prompt for a fresh Claude Code session. Activates when the user says "chat-to-code-prompt", "send to Claude Code", "make a CC prompt", "package this for Claude Code", "handoff to CC", or any variant. Produces one of three formats — task, plan, or context — chosen by the user at activation. Output is a single fenced markdown block at the end of the response so it can be tap-and-hold copied on mobile.

Designs, captures, and deploys Claude Code subagents (markdown files in .claude/agents/ or ~/.claude/agents/). Use when the user asks to design, create, scope, or build a subagent, lock a working chat pattern as a subagent, deploy an existing subagent file, convert a skill or slash command into a subagent, or references the /agents command, .claude/agents/ directory, or subagent frontmatter. Not for AGENTS.md, CLAUDE.md, slash commands, hooks, or general skills (use prompt-to-skill for skills).

Full-discovery planning skill that analyzes a GitHub repo, designs a complete implementation plan, and produces a structured markdown execution file for Claude Dispatch. Chains after interrogate. Activates when the user says "architect this", "plan for dispatch", "architect-plan-for-dispatch", "design this for dispatch", or when a large body of work needs thorough planning before handing off to Dispatch workers. Covers greenfield builds, refactors, migrations, infrastructure, documentation, and any complex multi-task project.

Covers effective use of the Apify CLI for web scraping and automation using Actors, datasets, key-value stores, and Crawlee. Covers apify init/run/push/login/call/pull, Actor concepts, pagination, proxy rotation, scheduling, webhooks, and CSV/JSON export. Activates when the user asks about Apify, running or building Actors, cloud web scraping, or extracting data at scale.

Covers effective use of bat, the syntax-highlighting cat replacement, for viewing files, previewing diffs, using as a MANPAGER, integrating with fzf previews, and customizing themes. Activates when the user asks about bat, syntax-highlighted file viewing, or replacing cat/less.

Covers effective use of DuckDB for in-process analytical SQL queries directly on CSV, Parquet, JSON, and remote S3/HTTP files without importing data. Activates when the user asks about DuckDB, querying files with SQL, analytical data processing, or OLAP on local files.

Covers effective use of fd as a modern, fast replacement for `find`. Covers fd PATTERN, -e ext, -t type, -H hidden, -I no-ignore, -x exec, -X batch exec, -E exclude, --changed-within, -d depth, --size, bulk operations, exec placeholders, and combining with ripgrep, bat, and fzf. Activates when the user asks about fd, finding files by name/type/date/size, or running commands on matched files.

Covers effective use of FFmpeg for video/audio conversion, trimming, filtering, streaming, and multimedia processing. Activates when the user wants to convert media files, trim video, create GIFs, extract audio, add subtitles, encode for web, or do any audio/video manipulation.

Covers effective use of fzf (fuzzy finder) for interactive filtering of files, command history, processes, git branches, and any line-oriented data. Activates when the user asks about fzf, interactive selection from lists, fuzzy searching, or filtering command output interactively.

Covers effective use of the GitHub CLI (gh) for managing pull requests, issues, releases, Actions workflows, repositories, and raw API calls. Activates when the user asks about gh, GitHub CLI, automating GitHub workflows, or interacting with GitHub from the terminal.

Covers effective use of HTTPie (http/https commands) for making HTTP requests from the terminal. Activates when the user asks about HTTPie, testing REST APIs from the command line, sending JSON bodies, managing auth tokens, or debugging HTTP requests interactively.

Covers effective use of ImageMagick for image conversion, resizing, cropping, compositing, batch processing, and creating favicons, thumbnails, watermarks, and animated GIFs. Activates when the user wants to manipulate images, convert formats, batch-resize photos, or create image assets programmatically.

Covers effective use of jq for processing, filtering, and transforming JSON data. Activates when the user asks about jq, parsing JSON in the terminal, reshaping API responses, or transforming structured data with filters, pipes, select, map, or reduce.

Covers effective use of lazygit for terminal UI git operations including staging, committing, pushing, branching, interactive rebase, merge conflict resolution, cherry-pick, bisect, and worktrees. Activates when the user asks about lazygit, TUI git workflows, or interactive git operations.

Covers effective use of Pandoc for converting between document formats including Markdown, HTML, DOCX, PDF, EPUB, LaTeX, and dozens more. Activates when the user wants to convert documents, generate PDFs from Markdown, create slide decks, manage citations, or batch-convert between wiki/documentation formats.

Covers effective use of Playwright for browser automation and testing. Covers CLI commands (codegen, test, show-report, screenshot, pdf), the core API (goto, click, fill, locator, waitForSelector, evaluate), and advanced patterns like visual regression, trace viewer, mobile emulation, network interception, and auth handling. Activates when the user asks about Playwright, browser automation, end-to-end testing, or web scraping with a real browser.

Covers effective use of ripgrep (rg) for fast regex search across codebases. Covers rg PATTERN, -i, -w, -l, -c, -n, --type, -g, -A/-B/-C, -U multiline, --json, type-add, .ripgreprc, replacement, and PCRE2. Activates when the user asks about ripgrep, rg, fast code search, or searching file contents with regex while respecting .gitignore.

Covers effective use of SQLite for serverless SQL databases including CLI dot-commands, schema management, CSV import/export, JSON functions, FTS5 full-text search, WAL mode, multi-database attach, and data analysis workflows. Activates when the user asks about SQLite, sqlite3, or embedded SQL databases.

Covers effective use of yq for reading, writing, and transforming YAML, JSON, XML, TOML, and CSV files. Activates when the user asks about yq, editing YAML configs, patching Kubernetes manifests, merging compose files, or converting between structured data formats.

Decision framework that evaluates any task against criteria (repetitiveness, judgment required, error tolerance, data sensitivity) and recommends whether to automate fully, AI-assist, or keep human. Activates when the user asks "should I automate this?", "delegate this?", "AI or human?", or when they're unsure whether a task belongs in an automation pipeline.

Expert architect-level review system that takes a software design document and systematically refines, reworks, and hardens it into a foolproof implementation blueprint with no blind spots. Runs a multi-phase pipeline: ingest, structural audit, technical deep dive, continuity check, gap analysis with autoheal, adversarial final review, and hardened output. Activates when the user says "review this design doc", "harden this design", "audit this spec", "review my architecture", "design doc review", or when the user provides a software design document and asks for feedback, critique, improvements, or validation. Also activate when the user uploads a technical spec, RFC, architecture doc, or system design and wants it improved, stress-tested, or made implementation-ready. Use this skill even when the user simply says "review this" and the input is clearly a software design document.

Stress-tests any idea, strategy, or position by attacking it from multiple angles until it's hardened with refined clarity. Default mode is rigorous sparring partner; escalates to full adversarial on request. Maintains a running hardening scorecard. Activates when the user says "devil's advocate", "stress test this", "attack this idea", "poke holes", "challenge this", "harden this", or when the user presents a position, strategy, or decision and asks for critical feedback or pushback.

Asks clarifying questions before executing any task. Activates automatically on ambiguous or complex requests, or manually when the user says "interrogate me", "ask me questions first", or "scope this". One question at a time, like a senior consultant scoping a project, until Claude is confident it fully understands the goal, constraints, audience, and hidden assumptions.

Translates a plain-English task description into a production-quality Claude prompt with structure, examples, and output formatting. Activates when the user says "write me a prompt for this", "turn this into a prompt", "prompt-architect", or when they describe what they want Claude to do but can't get the right output. Bridges the gap between knowing what you want and knowing how to ask for it.

Converts a working Claude prompt into a permanent, installable SKILL.md file with proper frontmatter, activation triggers, execution rules, and examples. Then automatically add it to the Summit Custom Skills table in the repo README and push to GitHub. Chains naturally after prompt-architect. Activates when the user says "make this a skill", "prompt to skill", "install this permanently", or when a prompt-architect output is confirmed working and the user wants to keep it.

Transforms Scribe AI (scribe.com) process documentation — step-by-step guides with screenshots and click annotations — into a complete custom software solution design document. Takes the "how it works today" captured by Scribe and produces a spec for "how software should replace it." Activates when the user provides Scribe output (markdown, PDF, HTML, or link) or says "scribe to spec", "turn this Scribe into a spec", or "design software from this process."

Converts a singular action, task, or one-off workflow into a modular system component that integrates with existing skills and systems. Analyzes the full inventory of available skills and system components, then runs a structured design dialog to define the component's interfaces, dependencies, data flow, and integration points. If no system exists yet, establishes a self-centric system architecture with the user at the center. Activates when the user says "systemize this", "make this a system", "turn this into a component", "system component", "build a system around this", or when the user has a working action that should become part of a larger interconnected system rather than remaining a standalone skill or one-off workflow. Also activate when the user describes wanting things to "work together", "connect", "feed into each other", or talks about building infrastructure around a recurring process.

Captures a successful one-off Claude interaction and convert it into a repeatable, templatized workflow with variables and trigger conditions. Activates when the user says "lock this workflow", "save this as a template", "I want to reuse this", or after a successful task where the user indicates they'll need to do it again. Turns one good result into infinite consistent results.

Create and initialize a new GitHub repository with sensible defaults using the gh CLI. Defaults to private. Sets up README, .gitignore, license, and .claude directory structure. Chains naturally into architect-plan-for-dispatch and any code-writing skill that needs a repo to target. Activates when the user says "create a repo", "new repo", "set up a GitHub repo", "gh-repo-create", or when another skill needs a fresh repository to write code into.

Analyze an application's source code and host hardware to recommend the smallest possible packaging strategy. Collects hardware specs via system commands, performs full dependency tree and CVE analysis, and produces packaging recommendations with generated Dockerfiles, CI/CD pipelines, benchmarking commands, and secret hygiene checks. Activates when the user says "pack light", "optimize this app", "shrink this", "smallest footprint", "analyze the application", or provides an application for deployment optimization.

Perform a comprehensive senior-architect-level code review of an entire codebase after a major update. Evaluates 10 dimensions (data integrity, test coverage, security, accessibility, performance, and more), assigns an overall letter grade, and produces a prioritized top-10 improvement plan with effort estimates. Activates when the user asks for a "code review", "architecture review", "codebase audit", "deep dive review", or "review this repo."

Coaching through Aristotle's published frameworks. Apply when the user needs virtue-based decision making, rhetorical skill, logical rigor, or the golden mean between extremes. Trigger with "ask Aristotle" or "Aristotle mode".

Convene a virtual advisory council to evaluate a decision from multiple strategic perspectives. Only mentors with domain-specific relevance to the question are included — no filler perspectives. Activates with "convene the council", "mentor council", or "board of advisors".