| name | ai-monetization-evaluation |
| description | Use when a user asks whether an AI project, agent, or workflow can autonomously make real money for them — including "AI 赚钱", "AI 被动收入", "AI 自动交易", "let AI make money for me", "AI side hustle", "passive income with AI", "AI 数字人赚钱", "AI 量化 自动下单". Evaluates claims through a 3-axis filter (real revenue path × regulatory friction × disclosure honesty) and produces a routed verdict (real-and-runnable / jurisdiction-blocked / research-grade / scam). Does NOT itself attempt to make money; identity-bound actions (registration, KYC, payments, real brokerage/exchange accounts) require explicit per-call user authorization. |
| triggers | ["AI 赚钱","AI 自动搞钱","AI 被动收入","AI side hustle","AI passive income","AI 自动交易","let AI make money for me","AI 量化","数字人 直播 赚钱","AI 副业","AI 一键 收入"] |
| category | research |
AI Monetization Claim Evaluation
The "AI makes money automatically" category is saturated with three things mixed together: real revenue paths, marketing fiction, and scams. This skill separates them before the user spends time, identity, or money.
When to use
- "Can this AI thing actually make me money?"
- "I want you to take over and earn for me"
- "Compare open-source projects that print revenue for the user"
- "Is this trading bot / agent / SaaS legit?"
- "Will the platform monetize my AI content?"
When NOT to use
- Real financial advice for a specific portfolio decision → refer to a licensed adviser
- Tax filing for specific income → refer to a CPA
- "Help me get rich quick" framing → refuse, route to scam-pattern literature
The 3-axis filter
Every candidate project must pass all three axes together:
Axis 1 ─ Real revenue path
□ Where does the money actually come from? (buyer, sponsor, exchange, ad share, commission)
□ Does the project move money on the user's behalf, or only produce content/research?
□ Is the revenue source reachable from the user's jurisdiction (KYC, age, country)?
Axis 2 ─ Regulatory friction
□ CFTC: trading automation, derivative signals, retail forex — which class are you in?
□ SEC: investment advice vs investment tool; stay in research-tool territory unless licensed
□ FTC: material-connection disclosure for affiliate content (#ad, #sponsored)
□ Platform AI-content policies (YouTube inauthentic, TikTok Creator Rewards, Meta)
□ Local tax reporting obligations (1099-K / 1099-MISC thresholds, VAT, GST, 境外所得)
Axis 3 ─ Disclosure honesty
□ README + LICENSE contain disclaimer hedges? "does not actually make trades",
"educational purposes only", "research only", "proof of concept",
"not investment advice", "no liability", "at your own risk"
□ Hedge present → research grade; not a real money-maker regardless of star count
□ Hedge absent → independently verify the execution path before trusting
A project passes only when all three check out. A single-axis failure is enough to disqualify.
Numbered evaluation procedure
gh repo view OWNER/REPO --json description,licenseInfo,stargazerCount,pushedAt,defaultBranchRef
rg -in 'order|trade|withdraw|publish|email|payout|register|deploy|charge|settle' README.md | head -40
rg -in 'educational|research|not advice|proof of concept|not financial|no liability|at your own risk' README.md LICENSE
rg -in 'binance|coinbase|alpaca|ibkr|stripe|paypal|adsense|affiliate|partnerstack' README.md | head -20
gh repo view OWNER/REPO --json pushedAt --jq '.pushedAt'
If step 2 produces only verbs like "analyzes / summarizes / generates", the project is research grade regardless of star count. Stars are a funding signal, not a money signal.
Decision routing (exactly one bucket)
Place the verdict in ONE of these buckets and tell the user which:
- Real revenue path, low friction → describe the path, then ask for explicit per-action authorization before any identity-bound step
- Real revenue path, jurisdiction-blocked → explain what unlocks it (KYC age, entity, US/EU residency) and what changes if the user qualifies
- Research tool only → useful as study/scaffolding; never call it a money-maker
- Marketing fiction / scam → name the red flags ("guaranteed returns", "passive income", "100x", zero-risk framing) and explain which regulator already calls it out
Refusal rules (absolute)
Apply before proposing any action:
- Never register an account, complete KYC, or submit identification documents in the user's name
- Never spend real money (gas, exchange funding, paid model API keys) without explicit per-call authorization
- Never connect to a brokerage / exchange / payment account the user has not personally authorized for that specific project
- Never promise income. "Guaranteed returns", "passive income", "set and forget", "AI prints money" are the literal CFTC red flags
- When the user says "you take over, make money" — clarify this skill produces a verdict, not a money-maker. Income still depends on the user's decisions, identity, and assets
Pitfalls
- Stars are a funding signal, not a money signal. A 90k-star "AI Hedge Fund" can still be marked "the system does not actually make any trades"
- "AI" prefix + money keyword ≠ income. Distinguish: AI analyzes → research, AI executes → candidate
- The same project may pass Axis 2 for one country and fail for another. Always ask jurisdiction before recommending
- "Staking yield" / "yield farming" / "liquidity mining" carry impermanent loss, slashing, counterparty risk — default to "no"
- Affiliate links require platform disclosure on most regimes (FTC, ASA, CMA). Disclosure absence is an independent disqualifier even when revenue is real
- The reason a project is open-source is often because it's research — research-grade work is published, production trading systems are not. Star count plus openness is often a tell
References
references/regulator-quickref.md — CFTC, SEC, FTC + platform AI-content policies, with the one rule each actually enforces and the source link
references/project-disclaimer-patterns.md — common README/LICENSE hedge phrasings and what each signals (POC, research, educational, no liability)
references/revenue-paths.md — concrete paths by category (trading, content, commerce, agent-as-service, data/API) with the typical KYC/age/jurisdiction snag each one hits