Claude Agent Stack

Visual content generation with Excalidraw hand-drawn aesthetic. USE WHEN user wants diagrams, visualizations, comics, or editorial illustrations.

Agent-driven memory via GCC commands. COMMIT progress, BRANCH explorations, MERGE results, CONTEXT to orient.

Use `just` to save and run project-specific commands. Use when the user mentions `justfile`, `recipe`, or needs a simple alternative to `make` for task automation.

Dynamic agent composition and management system. USE WHEN user says create custom agents, spin up custom agents, specialized agents, OR asks for agent personalities, available traits, agent voices. Handles custom agent creation, personality assignment, voice mapping, and parallel agent orchestration.

Adversarial iterative code review that loops until clean. USE WHEN user says "bugbot", "deep review", "adversarial review", "find all bugs", "review until clean", OR wants iterative bug hunting with test coverage.

CI/CD workflow enforcer that detects current dev stage and guides to the next step. USE WHEN user says "devflow", "what stage", "next step", "where am I", "check flow", "push to prod", "deploy", OR at any point during development to stay on track. Prevents deviations like working on main or editing prod directly.

Comprehensive research and content extraction — quick/standard/extensive/deep modes with multi-agent parallel research, content retrieval, AI trends analysis, and 242+ Fabric patterns. USE WHEN research, do research, quick research, extensive research, deep investigation, find information, investigate, extract alpha, analyze content, retrieve content, use fabric, AI trends, Claude research, enhance content, extract knowledge, interview research, web scraping, YouTube extraction, standard research.

Multi-mode analytical and creative thinking — first principles decomposition, iterative depth analysis, creative brainstorming, multi-agent council debates, adversarial red teaming, world threat modeling, and scientific hypothesis testing. USE WHEN first principles, decompose, deconstruct, reconstruct, challenge assumptions, iterative depth, multi-angle, deep exploration, be creative, brainstorm, divergent ideas, tree of thoughts, maximum creativity, technical creativity, idea generation, domain specific, council, debate, perspectives, quick consensus, red team, critique, stress test, adversarial validation, parallel analysis, devil's advocate, threat model, world model, future analysis, test idea, test investment, update models, view models, time horizon, think about, figure out, experiment, iterate, science, hypothesis, define goal, design experiment, quick diagnosis, structured investigation, full cycle.

Write and publish blog posts to yourdomain.com. USE WHEN "write a blog post" OR "blog about" OR "publish blog" OR "blog this work" OR "turn this into a blog post" OR "draft a post". Analyzes work context, generates Docusaurus MDX posts with hero images, deploys to GitHub Pages.

Code review and anti-pattern detection. USE WHEN user says review code, audit code, check for anti-patterns, code quality check, OR find bugs. Scans for security issues, anti-patterns, and project-specific violations.

Systematic UI exploration and documentation using headed Chrome (visible browser). USE WHEN documenting an interface into a spec, capturing a multi-step user flow, or exploring an unfamiliar UI to understand its structure. Run ChromeMCP first for authenticated sites. NOT for quick automation tasks (use AgentBrowser), NOT for writing scripts (use Browser), NOT just launching a browser (use ChromeMCP).

Create and validate skills. USE WHEN create skill, new skill, skill structure, canonicalize. SkillSearch('createskill') for docs.

Generates comprehensive specification documents from video demonstrations. USE WHEN user asks to "create spec from video" OR "analyze demo video" OR "document product from Loom/YouTube". Navigates video, captures screenshots, and writes detailed feature documentation.

Automates browser interactions for web testing, form filling, screenshots, and data extraction. USE WHEN user needs to navigate websites OR interact with web pages OR fill forms OR take screenshots OR test web applications OR extract information from web pages.

Observable browser automation using Chrome MCP tools. Use when you need to browse websites, take screenshots, interact with web pages, or perform browser tasks in your current Chrome. Keywords - browse, screenshot, browser, chrome, bowser, ui testing, observable.

Social media scraping, business data, e-commerce via Apify actors — with auto-update workflow for actor catalog. USE WHEN Twitter, Instagram, LinkedIn, TikTok, YouTube, Facebook, Google Maps, Amazon scraping, Apify, update Apify actors, social media scraping, lead generation, web scraper.

Network reconnaissance — subdomain enumeration, port scanning, DNS/WHOIS/ASN lookups, endpoint discovery from JS, mass scanning, path discovery, CIDR/netblock analysis. Passive and active modes with corporate structure mapping and bounty program tracking. USE WHEN recon, reconnaissance, bug bounty, bounty programs, attack surface, subdomains, subdomain enumeration, port scan, mass scan, DNS, WHOIS, ASN, CIDR, netblock, IP recon, domain recon, passive recon, corporate structure, endpoint discovery, path discovery, scan results, analyze scan, update tools.

Objective eval metrics via code/model/human graders with pass@k/pass^k scoring. USE WHEN eval, evaluate, test agent, benchmark, verify behavior, regression test, capability test, run eval, compare models, compare prompts, create judge, create use case, view results, failure to task, suite manager, transcript capture, trial runner.

Audit and optimize agent context files across Claude Code and OpenAI Codex. Covers CLAUDE.md, AGENTS.md, rules, skills, and memory organization. USE WHEN "optimize context" OR "audit setup" OR "context architecture" OR "where should this go" OR "reduce duplication" OR "context budget" OR "too many tokens" OR "reorganize rules" OR "skill hierarchy" OR "CLAUDE.md too big" OR "AGENTS.md too big" OR "optimize skills". Based on Gloaguen et al. 2026 research, Anthropic best practices, and OpenAI Codex docs.

Optimizes CLAUDE.md, AGENTS.md, and other agent context files in a repository using research-backed principles. USE WHEN optimize context files OR audit CLAUDE.md OR improve agents.md OR reduce context bloat OR review agent instructions. Based on "Evaluating AGENTS.md" (Gloaguen et al., 2025) findings.

Multi-agent debate with visible transcripts where agents respond to each other. USE WHEN council, debate, perspectives, weigh options, deliberate, multiple viewpoints. Unlike RedTeam (adversarial), Council is collaborative-adversarial.

32 adversarial agents to destroy weak arguments and find fatal flaws — parallel analysis and adversarial validation. USE WHEN red team, attack idea, counterarguments, critique, stress test, poke holes, devil's advocate, find weaknesses, break this, parallel analysis, adversarial validation.

Hypothesis-test-analyze cycles for systematic problem-solving — the meta-skill governing all others. Includes define goal, generate hypotheses, design experiment, measure results, analyze results, iterate, full cycle, quick diagnosis, and structured investigation. USE WHEN think about, figure out, try approaches, experiment with, iterate on, improve, optimize, define goal, generate hypotheses, design experiment, measure results, analyze results, full cycle, quick diagnosis, structured investigation, science, hypothesis.

240+ specialized prompt patterns for content analysis, extraction, and transformation — run patterns, sync from repo, create threat models, summarize. USE WHEN fabric, fabric pattern, run fabric, update patterns, sync fabric, summarize with fabric, create threat model, analyze with fabric.

Extract structured JSON from URLs, files, videos, PDFs with entity extraction and batch support. USE WHEN parse, extract, URL, transcript, entities, JSON, batch, content, YouTube, PDF, article, newsletter, Twitter, browser extension, collision detection, detect content type, extract article, extract newsletter, extract YouTube, extract PDF, parse content.

Headless browser automation using Playwright CLI. Use when you need headless browsing, parallel browser sessions, UI testing, screenshots, web scraping, or browser automation that can run in the background. Keywords - playwright, headless, browser, test, screenshot, scrape, parallel.

Visual verification and browser automation via Playwright. Headless or headed Chrome. USE WHEN browser, screenshot, debug web, verify UI, troubleshoot frontend, automate browser, browse website, review stories, run stories, recipe, web automation.

Structured OSINT investigations — people lookup, company intel, investment due diligence, entity/threat intel, domain recon, organization research using public sources with ethical authorization framework. USE WHEN OSINT, due diligence, background check, research person, company intel, investigate, company lookup, domain lookup, entity lookup, organization lookup, threat intel, discover OSINT sources.

4-tier progressive URL scraping and multi-page crawling — WebFetch, then Chrome-header curl, then Playwright browser, then Bright Data proxy. Auto-escalates when lower tiers fail. USE WHEN Bright Data, scrape URL, web scraping, site blocking me, can't access, bot detection, crawl site, crawl pages, spider, CAPTCHA, four tier scrape, progressive scraping, Chrome headers.

Security news from tldrsec, no.security, Krebs, Schneier, and other sources. USE WHEN security news, security updates, what's new in security, breaches, security research, sec updates, tldrsec, Krebs, Schneier.

Divergent ideation via Verbalized Sampling + extended thinking (1.6-2.1x diversity). USE WHEN be creative, deep thinking, brainstorm, divergent ideas, creative solutions, maximum creativity, tree of thoughts, idea generation, domain specific creativity, technical creativity, standard creativity.

CRUD on {PRINCIPAL.NAME}'s curated aphorism collection — search by theme, add with metadata, research thinkers, match quotes to newsletter content. USE WHEN aphorism, quote, saying, find quote for newsletter, research thinker quotes, add aphorism, search aphorisms, find aphorism.

Generate production-ready TypeScript CLIs with 3-tier template system (simple argv, yargs, oclif), type safety, error handling, and documentation. USE WHEN create CLI, build CLI, command-line tool, wrap API, add command, upgrade tier, TypeScript CLI.

Deploy and manage Cloudflare Workers, Pages, and services via Code Mode MCP (API queries) + wrangler (deploys). OAuth auth for wrangler (tokens lack Pages perms). USE WHEN Cloudflare, worker, deploy, Pages, MCP server, wrangler, DNS, KV, R2, D1, Vectorize.

Parallelize work via background/foreground agents, built-in types, custom agents, or agent teams/swarms. USE WHEN 3+ independent workstreams, parallel execution, agent specialization, Extended+ effort, agent team, swarm, create an agent team.

Extract system improvements from content AND monitor external sources (Anthropic ecosystem, YouTube). USE WHEN upgrade, improve system, system upgrade, analyze for improvements, check Anthropic, Anthropic changes, new Claude features, check YouTube, new videos, algorithm upgrade, mine reflections, find sources, research upgrade, PAI upgrade.

Aggregate and analyze annual security reports from major vendors — extract trends, compare threat landscapes year-over-year, produce synthesized threat intelligence summaries. Fetch, list, and update report sources. USE WHEN annual reports, security reports, threat reports, industry reports, vendor reports, threat landscape, fetch report, list sources, update sources.

Full web app security assessment — app understanding, threat modeling, OWASP testing, ffuf fuzzing, Playwright automation, AI-assisted vuln analysis. Coordinates with Recon and PromptInjection skills. USE WHEN web assessment, pentest, security testing, vulnerability scan, threat model app, understand application, create threat model, vulnerability analysis, Gemini analysis, ffuf, fuzzing, bug bounty, OSINT, browser automation, Playwright.

2-8 scientific lens passes to surface hidden requirements single-pass analysis misses. USE WHEN iterative depth, deep exploration, multi-angle analysis, multiple perspectives, examine from angles, surface hidden requirements.

Create, merge, split, extract text/tables from PDFs, fill forms, add watermarks, and convert to/from other formats. USE WHEN pdf, PDF file, merge PDF, extract tables, fill form, split PDF.

Decompose to axioms, challenge inherited assumptions, reconstruct from verified truths. USE WHEN first principles, fundamental, root cause, decompose, challenge assumptions, rebuild from scratch.

Create, edit, and analyze Word documents — tracked changes, tables, images, complex formatting, and format conversion. USE WHEN docx, Word document, create document, tracked changes, redlining.

Create, edit, and analyze PowerPoint presentations — slide layouts, speaker notes, design elements, and format conversion. USE WHEN pptx, PowerPoint, slides, presentation, create slides, speaker notes.

Content-adaptive wisdom extraction — detects what domains exist in content and builds custom sections (not static IDEAS/QUOTES). Produces tailored insight reports from videos, podcasts, articles. USE WHEN extract wisdom, analyze video, analyze podcast, extract insights, what's interesting, extract from YouTube, what did I miss, key takeaways.

Stress-test ideas, strategies, and investments across 11 time horizons (6mo-50yr). Update and view world models. USE WHEN threat model, world model, test idea, test strategy, future analysis, test investment, test against future, stress test idea, time horizon analysis, update models, view models, refresh models, model status.

Ethical people-finding using 15 parallel research agents (45 search threads) across public records, social media, reverse lookups. Public data only, no pretexting. USE WHEN find person, locate, reconnect, people search, skip trace, reverse lookup, social media search, public records search, verify identity.