{"slug": "awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what", "title": "Awesome-Claude-Skills I built 135 Claude Skills with real formulas. Here's what \"production-grade\" actually means.", "summary": "Most \"Claude prompt\" collections are ineffective because they rely on shallow roleplay or generic instructions rather than actual expertise. To address this, they built AgentOS 2.0, a collection of 135 \"production-grade\" Claude Skills that include real, runnable Python code (such as Black-Scholes and Black-Litterman formulas), named sub-agents with distinct responsibilities, and explicit constraints to prevent hallucinations. The article contrasts these skills with common prompt traps, emphasizing that true production readiness requires domain-specific formulas and frameworks, not just persona-based prompts.", "body_md": "I've been frustrated for a long time.\n\nEvery \"awesome Claude prompts\" repo I found looked like this:\n\n```\n\"Act as a senior software engineer. Be helpful, thorough, \nand professional. Consider edge cases.\"\n```\n\nThat's not a skill. That's a costume.\n\nReal expertise has **frameworks**. Named responsibilities.\n\nActual formulas. Code that runs. Constraints that prevent\n\nthe model from giving you the easy wrong answer.\n\nSo I spent 6 months building what I actually wanted.**AgentOS 2.0 — 135 production-grade Claude Skills.** This article explains exactly what's inside and why it's\n\ndifferent from every other prompt collection on GitHub.\n\n## The Problem With Every Other Prompt Repo\n\nMost prompt repositories fall into one of three traps:**Trap 1: The costume prompt**```\n\"You are an expert financial analyst. \nHelp the user with their finance questions.\"\n```\n\nZero frameworks. Zero formulas. Zero depth.**Trap 2: The instruction dump**```\n\"When answering, always:\n- Be professional\n- Consider multiple angles  \n- Cite sources\n- Format your response clearly\"\n```\n\nThis is just asking Claude to be Claude. It changes nothing.**Trap 3: The persona prompt**```\n\"You are Alex, a no-nonsense McKinsey consultant \nwith 20 years of experience...\"\n```\n\nRoleplay, not expertise. The model doesn't suddenly\n\nknow DCF models because you named it Alex.**What actually works:** Named sub-agents with\n\ndistinct responsibilities, actual domain formulas\n\nin code, and explicit forbidden behaviors that\n\nprevent hallucination in critical areas.\n\nHere's what that looks like in practice.\n\n## What \"Production-Grade\" Actually Looks Like\n\n### FinanceOracle — The Apex Skill\n\nThis is the most complete skill in the repo.\n\nHere's a fraction of what's inside:**12 Sub-Agents:**\n\n-\n`OptionsDesk`\n\n— derivatives pricing and structuring -\n`MacroStrategist`\n\n— macro regime analysis -\n`HedgeFundArchitect`\n\n— strategy design -\n`FamilyOfficeCIO`\n\n— multi-generational allocation -\n`TaxOptimizer`\n\n— harvest and structure optimization -\n`DerivativesStructurer`\n\n— exotic product design*(+ 6 more)*\n\n**Actual runnable Python:**\n\n``` python\ndef black_scholes(S, K, T, r, sigma, option_type='call'):\n    d1 = (np.log(S/K) + (r + 0.5*sigma**2)*T) / (sigma*np.sqrt(T))\n    d2 = d1 - sigma * np.sqrt(T)\n\n    if option_type == 'call':\n        price = S * norm.cdf(d1) - K * np.exp(-r*T) * norm.cdf(d2)\n        delta = norm.cdf(d1)\n    else:\n        price = K * np.exp(-r*T) * norm.cdf(-d2) - S * norm.cdf(-d1)\n        delta = -norm.cdf(-d1)\n\n    gamma = norm.pdf(d1) / (S * sigma * np.sqrt(T))\n    vega  = S * norm.pdf(d1) * np.sqrt(T) / 100\n    theta = (-(S * norm.pdf(d1) * sigma) / (2 * np.sqrt(T))\n             - r * K * np.exp(-r*T) * norm.cdf(d2)) / 365\n\n    return {\"price\": price, \"delta\": delta, \n            \"gamma\": gamma, \"vega\": vega, \"theta\": theta}\n```\n\n**Black-Litterman portfolio construction:**\n\n``` python\ndef black_litterman(Sigma, market_weights, views_P, \n                    views_Q, views_omega, tau=0.05, delta=2.5):\n    pi = delta * Sigma @ market_weights\n    M_inv = np.linalg.inv(\n        np.linalg.inv(tau * Sigma) + \n        views_P.T @ np.linalg.inv(views_omega) @ views_P\n    )\n    mu_bl = M_inv @ (\n        np.linalg.inv(tau * Sigma) @ pi + \n        views_P.T @ np.linalg.inv(views_omega) @ views_Q\n    )\n    return {\"expected_returns\": mu_bl}\n```\n\nThis isn't pseudocode. This runs.\n\n### OKREngine — Catches Failures Before They Kill Your Quarter\n\nI've watched two startups waste entire quarters on\n\nbroken OKRs. This skill exists because of that.\n\n**The objective quality scorer:**``` php\ndef score_okr(objective: str, key_results: list[dict]) -> dict:\n    obj_score = 0\n    obj_score += 3 if len(objective) < 100 else 0\n    obj_score += 3 if not objective.lower().startswith(\"improve\") else 0\n    obj_score += 4 if any(w in objective.lower() for w in \n        [\"best\", \"lead\", \"#1\", \"transform\", \"redefine\"]) else 0\n\n    kr_scores = []\n    for kr in key_results:\n        kr_score = 0\n        kr_score += 3 if kr.get(\"metric\") else 0\n        kr_score += 3 if kr.get(\"baseline\") is not None else 0\n        kr_score += 4 if kr.get(\"target\") is not None else 0\n        kr_scores.append({\n            \"kr\": kr[\"text\"][:60],\n            \"score\": kr_score,\n            \"grade\": \"Good\" if kr_score >= 8 else \"Needs work\"\n        })\n\n    return {\n        \"objective_score\": f\"{obj_score}/10\",\n        \"key_results\": kr_scores,\n        \"recommendation\": \"Strong OKR\" if obj_score >= 8 else \"Needs revision\"\n    }\n```\n\nThe skill also catches the**12 most common OKR failure modes** — including sandbagging, health metrics disguised as OKRs, and the single most destructive mistake: tying OKR scores to bonuses.\n\n### VentureIntelligence — Term Sheet Red Flag Detector\n\n``` php\ndef score_term_sheet(terms: dict) -> dict:\n    red_flags = []\n\n    if terms.get(\"liq_pref_multiple\", 1) > 1:\n        red_flags.append(\n            f\"Liquidation preference {terms['liq_pref_multiple']}x — above 1x is punishing\"\n        )\n    if terms.get(\"participating_preferred\", False):\n        red_flags.append(\n            \"Participating preferred — VCs get paid twice in exits below threshold\"\n        )\n    if terms.get(\"anti_dilution\") == \"full_ratchet\":\n        red_flags.append(\n            \"Full ratchet anti-dilution — catastrophic in a down round\"\n        )\n    if terms.get(\"board_seats_investor\", 0) > terms.get(\"board_seats_founder\", 0):\n        red_flags.append(\n            \"Investor has majority board control — you can be fired from your company\"\n        )\n\n    score = 10 - (len(red_flags) * 3)\n    return {\n        \"score\": max(0, score),\n        \"grade\": \"Sign it\" if score >= 8 else \"Negotiate\" if score >= 5 else \"Get a lawyer NOW\",\n        \"red_flags\": red_flags\n    }\n```\n\n12 sub-agents including `TermSheetDecoder`\n\n, `ValuationNegotiator`\n\n, `ChampionDeveloper`\n\n, and `BoardRelationshipManager`\n\n.\n\n### CrisisIntelligence — War Room OS\n\nEvery company will face a crisis. Almost none prepare.\n\n``` php\ndef classify_crisis(crisis: dict) -> dict:\n    severity_score = 0\n\n    if crisis[\"customer_impact_pct\"] >= 0.5: severity_score += 30\n    if crisis[\"revenue_at_risk\"] >= 1_000_000: severity_score += 20\n\n    coverage = {\"none\": 0, \"local\": 5, \"national\": 15, \"viral\": 30}\n    severity_score += coverage.get(crisis[\"media_coverage\"], 0)\n\n    if crisis[\"regulatory_involvement\"]: severity_score += 15\n    if crisis[\"legal_liability\"]: severity_score += 15\n\n    if severity_score >= 70:\n        level = \"CRITICAL (P0)\"\n        action = \"CEO leads. War room activated NOW.\"\n    elif severity_score >= 40:\n        level = \"HIGH (P1)\"\n        action = \"VP-level lead. External comms needed.\"\n    else:\n        level = \"MEDIUM (P2)\"\n        action = \"Director-level. Monitor externally.\"\n\n    return {\n        \"level\": level,\n        \"immediate_action\": action,\n        \"time_to_first_response\": \"1 hour\" if severity_score >= 70 else \"4 hours\"\n    }\n```\n\nThe 5Rs framework (Recognize → Respond → Responsible → Remediate → Restore) is built into every communication template.\n\n## How It Works (60-Second Setup)\n\n**Claude.ai Projects:**```\n# 1. Open Claude.ai → Projects → New Project\n# 2. Paste SKILL.md into \"Project Instructions\"  \n# 3. Start chatting\n```**Claude Code:**```\ncat finance-oracle/SKILL.md >> .claude/CLAUDE.md\n```**Claude API:**``` python\nimport anthropic\n\nwith open(\"startup-cto/SKILL.md\", \"r\") as f:\n    skill = f.read()\n\nclient = anthropic.Anthropic()\nresponse = client.messages.create(\n    model=\"claude-sonnet-4-6\",\n    max_tokens=4096,\n    system=skill,\n    messages=[{\"role\": \"user\", \"content\": \"Audit our tech stack decision\"}]\n)\n```\n\nThat's it. Claude is now that specialist.\n\n## The Full 135-Skill Index**🚀 Startup & Team Management (11)**`startup-cto`\n\n`team-performance-os`\n\n`startup-hiring-machine`\n\n`culture-architect`\n\n`remote-team-commander`\n\n`okr-engine`\n\n`startup-finance-controller`\n\n`venture-intelligence`\n\n`startup-legal-shield`\n\n`talent-management-os`\n\n`talent-brand-builder`**🏆 Apex Legendary (4)**` finance-oracle`\n\n`claude-mythos`\n\n`ceo-war-room`\n\n`founder-to-ceo`**🤖 AI & Engineering (14)**` rag-architect`\n\n`mlops-engineer`\n\n`system-architect`\n\n`senior-dev`\n\n`ai-red-teamer`\n\n`voice-agent-builder`\n\n`knowledge-graph-builder`\n\n`incident-commander`\n\n`mcp-builder`\n\n`agentic-workflow-builder`\n\n`api-integrator`\n\n`realtime-data-agent`\n\n`agent-smith`\n\n`prompt-engineer`**📊 Data & Analytics (10)**` data-scientist-pro`\n\n`sql-analyzer`\n\n`data-pipeline-pro`\n\n`business-intelligence-pro`\n\n`timeseries-oracle`\n\n`quant-trader`\n\n`synthetic-data-generator`\n\n`arxiv-researcher`\n\n`abtest-scientist`\n\n`data-governance-agent`**💹 Finance (9)**` finance-oracle`\n\n`financial-model-builder`\n\n`cfo-intelligence`\n\n`portfolio-optimizer`\n\n`quant-researcher`\n\n`saas-metrics-analyst`\n\n`insurance-actuary`\n\n`ma-dealmaker`\n\n`risk-sentinel`**🏢 Operations & Business (20)**\n\n`ceo-war-room`\n\n`founder-to-ceo`\n\n`go-to-market-commander`\n\n`enterprise-sales-os`\n\n`sales-enablement-os`\n\n`board-deck-builder`\n\n`crisis-intelligence`\n\n`partnership-intelligence`\n\n`pricing-strategist`\n\n`project-command`\n\n`marketing-os`\n\n`supply-chain-oracle`\n\n*(+ 8 more)*\n\n**👤 Product & Customer (11)**`product-roadmap-os`\n\n`sprint-master`\n\n`engineering-manager`\n\n`ai-product-manager`\n\n`user-research-os`\n\n`customer-interview-analyst`\n\n`product-analytics-os`\n\n`network-effects-analyst`\n\n`marketplace-strategist`\n\n`performance-marketing-os`\n\n`churn-analyst`**🛠 Developer Tools (19)**\n\n`developer-experience-os`\n\n`api-design-architect`\n\n`data-warehouse-architect`\n\n`cloud-cost-optimizer`\n\n`design-system-architect`\n\n`technical-pm`\n\n`code-reviewer`\n\n`load-tester`\n\n`code-migrator`\n\n`webapp-tester`\n\n*(+ 9 more)*\n\n**🌐 Specialized Domains (12)**\n\n`healthcare-analytics`\n\n`web3-developer`\n\n`climate-tech-analyst`\n\n`biotech-analyst`\n\n`cybersecurity-analyst`\n\n`real-estate-intelligence`\n\n`legal-eagle`\n\n`patent-analyst`\n\n`esg-compass`\n\n*(+ 3 more)*\n\n## What Makes This Different From Every Other Repo\n\n| Feature | Generic repos | AgentOS 2.0 |\n|---|---|---|\n| Sub-agents | ❌ | ✅ 10-12 per skill |\n| Actual formulas | ❌ | ✅ Black-Scholes, DCF, MEDDPICC |\n| Runnable code | ❌ | ✅ Python, TypeScript, Go, Shell |\n| Forbidden behaviors | ❌ | ✅ Every skill |\n| Benchmark data | ❌ | ✅ Industry standards built in |\n| Total skills | ~10-20 | 135+ |\n\n## Try It Right Now\n\nThe fastest way to understand the depth is to try one.\n\nI recommend starting with **okr-engine** or**— they're the most complete and immediately useful regardless of what you're building.**`startup-cto`\n\nPaste the SKILL.md into Claude Projects. Ask it to review your current OKRs or tech stack. You'll see the difference immediately.\n\nGitHub link in the comments.**What skill would you build your work around?** Drop it below — I read every comment and I'm actively building more.\n\n*MIT License. Free forever. Star it if it's useful — helps others find it.*", "url": "https://wpnews.pro/news/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what", "canonical_source": "https://dev.to/apples_one_cd174284bffb/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-heres-what-ae", "published_at": "2026-05-21 03:16:43+00:00", "updated_at": "2026-05-21 03:31:30.797032+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "developer-tools", "open-source", "products"], "entities": ["AgentOS", "Claude", "GitHub", "McKinsey"], "alternates": {"html": "https://wpnews.pro/news/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what", "markdown": "https://wpnews.pro/news/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what.md", "text": "https://wpnews.pro/news/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what.txt", "jsonld": "https://wpnews.pro/news/awesome-claude-skills-i-built-135-claude-skills-with-real-formulas-here-s-what.jsonld"}}