Inspiration
I'm a UGC creator and brand contracts are where creators quietly lose money. But the more I looked at it, the more I realized this isn't a creator problem. It's an everyone-who-signs-contracts problem, and creators are just the clearest example to me personally. The fee looks fine on page one, then page four says Net 45 from invoice receipt, perpetual usage rights and a view bonus that works out to half my rate. I know my rates. What I didn't have was a way to check every contract against them before signing, especially when deals pile up during a busy month. Talent managers have the same problem multiplied by every creator on their roster.
What it does
Rate Card Guard checks every brand or client contract on your NAS against your own rate card.
You save your terms once: minimum fee per post, rate per 100K views, payment schedule, deposit, usage rights, paid ads, exclusivity, revision rounds, kill fee and any hard no terms. Then you say "check my Brand Deals folder against my rate card" and get a scorecard for each contract. Every term is marked Meets, Falls short, Missing or Needs review, with the exact clause quoted and its page number, plus a counter ask you can send back to the brand.
It also does the math creators skip. A brand offering $0.80 per 1,000 views is offering $80 per 100K, and if your rate is $150 the scorecard says so. Talent managers can keep one rate card per creator and check a whole batch at once. When several contracts are checked, it ranks them from best to worst fit.
Your contracts never leave the NAS. No external AI service is used and original files are never moved, renamed or edited.
Beyond creators
A rate card is really just a list of terms you won't go below. Every industry that signs contracts has one, written down or not, and almost nobody checks every incoming contract against it. Rate Card Guard works the same way for any of them: swap the rate card for that industry's standard terms and the Skill finds, extracts, compares and quotes in exactly the same way.
| Who | Their "rate card" | What it catches |
|---|---|---|
| Talent managers and agencies | One rate card per creator on the roster | A batch of brand deals checked against the right creator's terms |
| Freelancers and consultants | Day rate, deposit, net terms, IP ownership | SOWs with Net 60, no deposit or full IP transfer before payment |
| Photographers and videographers | Licensing scope, usage period, kill fee | Buyouts disguised as single-use licenses |
| Musicians and performers | Guarantee, deposit, cancellation terms, rider | Booking contracts with no cancellation fee or a missing deposit |
| Event planners and venues | Minimums, cancellation windows, liability limits | Vendor contracts that shift all cancellation risk onto them |
| Small business owners | Standard vendor and client terms | Auto-renewals, uncapped liability and payment terms longer than their cash flow allows |
| Property managers | Lease standards | Leases that drift from the standard on deposits, renewals or maintenance |
Anyone who signs more contracts than they have time to read closely has this problem, and a NAS is where those contracts already live.
How we built it
Rate Card Guard is a Markdown-only Skill built with the UGOS Skill Creation Tool and the Uliya Skill Development Standard. The Skill orchestrates three NAS experts:
- file-agent finds contracts in the folder you name, reads your saved rate card and optionally tags contracts that fall short
- document-agent extracts deal terms with the contract schema (including OCR for scanned contracts) and then locates and quotes the clause behind every flag
- knowledge-agent saves your rate card and finished scorecards back to the NAS
Each step passes only verified upstream results to the next one. Scoring follows deterministic comparison rules in a linked reference, covering unit conversion, payment clock checks and hard no terms. Reads run without interruption, while saving and tagging wait for platform approval. Every delegation state (completed, failed, cancelled, waiting_user and waiting_approval) has explicit handling.
Challenges we ran into
I actually started with a different idea: a personal archive Skill that turned saved screenshots and footage into lessons and content ideas. When I mapped it against the capability catalog, the NAS could read screenshots well but couldn't visually analyze video, which was half the concept. Rather than submit something built on workarounds, I pivoted on deadline day to an idea where the NAS document tools are the core strength instead of the weak spot.
Writing a Skill is also different from writing an app. There's no code to lean on, so every instruction has to be self-contained and precise enough that an expert agent can't misread it. The hardest part was making sure the Skill never guesses a contract term, never invents a file path and treats a missing clause differently from an unreadable one.
Accomplishments that we're proud of
- It passes the official kit checker with zero errors and zero warnings
- Every flag is grounded in a quoted clause and a page number, so nothing in the scorecard is unsupported
- It handles the details that actually cost creators money: view rate conversions, when the payment clock starts and perpetual rights hidden in usage clauses
- It works for a solo creator and scales to a talent manager reviewing contracts for a whole roster
What we learned
I learned to design around what a platform truly supports instead of what I wish it supported. Reading the capability catalog closely changed my whole project, and the result is stronger for it. I also learned how much care goes into agent orchestration: clear inputs, honest failure states and approval before anything changes on someone's storage.
What's next for Rate Card Guard
- Testing on a real UGREEN AI NAS with synthetic contracts, including scanned and password protected files
- Ready-made rate card templates for other industries: freelance SOWs, photography licensing, performance bookings, vendor agreements and leases
- Watching a contracts folder and scoring new ones as they arrive
- Matching contracts against invoices and payment confirmations to show what was billed versus what was paid
- Tracking terms over time so users can see their deals improve
Built With
- agent-orchestration
- ai-agents
- claude
- document-ai
- markdown
- ocr
- skills-hub
- ugos
- ugreen-ai-nas
- uliya-skill-standard