Verifying Before Shipping: A RBNT Token Case Study

How checking every claim, instead of trusting the brief, shaped a Redbelly DAO submission.

When I picked up TASK-11 on the Redbelly DAO task board, the brief looked simple enough on paper. Write a report explaining what RBNT actually does. Submit RBNT to DeFiLlama for TVL tracking. Get Redbelly listed on RWA.xyz. Write a short explainer for the community. Four deliverables, one task, done in a weekend.

It did not go that way. What actually happened was a slow process of discovering that two of the four assumptions baked into the task brief were wrong, and that the only way to find that out was to check everything myself instead of taking the brief at face value. This is the story of what I built, what I found, and why the verification mattered more than the deliverables themselves.

Part 1: What RBNT is actually for

The first deliverable was a tokenomics report. Easy enough to write badly. Anyone can summarize a whitepaper in a paragraph and call it done. I wanted something a reader could actually trust, so every number in the report traces back to Redbelly's official whitepaper from August 2025, not to a secondary source or a guess.

RBNT has a fixed supply of ten billion tokens. No inflation, no ongoing minting, no burn mechanism written into the protocol. It does five real jobs on the network: paying gas, gating governance votes, backing three separate staking flavors (consensus, SEVM, and oracle), locking for shard participation, and funding incentives for node operators and ecosystem builders.

The allocation table breaks down like this:

Release is staggered rather than a single cliff. The Governance DAO allocation starts unlocking in month one. Team tokens do not start releasing until month twelve, and only while the recipient is actively engaged with Redbelly. Most buckets stretch out to month 32 through 36. Ecosystem Development and Reserve do not follow a fixed schedule at all. They release at the network's discretion.

One detail worth calling out: Redbelly prices gas at a flat rate roughly a cent per standard transfer in dollar terms regardless of what RBNT itself is doing in the market. For a chain built around real-world assets, that is a deliberate choice. Institutions moving real assets need predictable transaction costs, not fees that spike with speculation.

The report also includes a four-way competitor comparison against Ondo, Polymesh, and XDC, and it explicitly avoids making any price prediction. The whitepaper does not publish a valuation model, so neither did I. RBNT's value proposition is in what it lets the network do, not in a number someone assigns it later.

Part 2: The bug that was hiding in plain sight

The second deliverable asked for a DeFiLlama TVL adapter for reddex, Redbelly's native decentralized exchange, built against DeFiLlama's public SDK. I built the adapter, wrote local mock tests against it, and then wrote a manual verification script to sanity check the numbers before submitting anything.

That manual script is where things got interesting. It kept returning a 404 error every time it tried to pull live price data. On inspection, the bug was a single wrong hostname. DeFiLlama splits its API across two different hosts: one domain handles TVL and protocol data, and a completely separate domain handles price data. My script was hitting the TVL domain for a price route that simply does not exist there. A one-line fix sed-swapping the hostname resolved it immediately.

What made this worth writing up rather than just quietly fixing was where else the same mistake was hiding. The identical wrong hostname showed up in three more places across the submission kit: the bundled copy of the verification script, the testing documentation, and the API endpoint reference docs. The actual production adapter code was unaffected, because it relies on DeFiLlama's SDK internally, which handles pricing correctly on its own. But the manual verification tooling meant to double-check the SDK's work had the bug baked into every copy of itself.

I fixed all four instances and added an explicit note in the API documentation flagging that price data lives on a separate host, specifically so the next person onboarding a protocol does not repeat the same mistake.

Once the script actually worked, I ran the real comparison. My manual on-chain calculation of Reddex's TVL on the Redbelly chain came out to $21,778.26. DeFiLlama's own live figure for the same pool was $22,002.09. That is a 1.02% difference comfortably inside the 5% margin the task required for a passing verification.

There was one more honest gap worth flagging rather than hiding. Two of the six pooled tokens I checked including a wrapped native pairing token that appears across three different pairs had no price data available on DeFiLlama's price endpoint. That means my manual figure is a slight undercount even after the fix, though it still lands well within the accuracy threshold. I documented this explicitly in the testing notes rather than smoothing it over. A verification report that hides its own edge cases is not actually a verification report.

Part 3: When the brief itself was wrong

The third deliverable was where the task brief turned out to be outdated in a way that mattered. The original instructions assumed Redbelly was absent from both DeFiLlama and RWA.xyz, and asked for submission kits for both. By the time I actually checked DeFiLlama directly, Redbelly was already listed there meaning the premise behind part of the brief no longer held. That is a distinct kind of finding from just filling out a form. It means the first job was not submission; it was checking whether submission was even still necessary, and documenting what changed.

RWA.xyz was a different story. I cross-referenced Redbelly's status directly against RWA.xyz's own published directory and their network coverage list, which spans 26 networks. Redbelly was not on it. That absence was genuine, not outdated information.

There was a wrinkle worth flagging here too. A domain called rwa.io (without the "xyz") already had a Redbelly profile. That is a different company and a different product from RWA.xyz, despite the nearly identical name. It does not count toward RWA.xyz coverage, and I flagged this explicitly in the submission kit so nobody on the review side gets confused by the naming collision.

RWA.xyz also does not run a self-service submission process the way DeFiLlama does. There is no pull request to open. Their onboarding runs through a Partners App signup, a Slack channel, a kickoff call, and a data collection phase before any listing goes live. So instead of writing code, this part of the task became a documentation exercise:

• An onboarding checklist mapping their actual seven-step process
• A network profile draft with every field pre-filled including a note flagging the difference between Redbelly's mainnet chain ID (151) and its testnet chain ID (153) to prevent a mixup during review
• A list of three sourced ecosystem partners worth introducing to RWA.xyz
• A template for the three points of contact their kickoff call typically asks for

Part 4: Explaining it without losing anyone

The fourth deliverable was a short explainer under 500 words meant for Discord and social sharing. The hard part of writing something short is not fitting the words. It is deciding what to cut without cutting anything that matters.

The final version runs 419 words and keeps the same sourcing discipline as the full report: the five real uses, the full allocation table, a summary of the vesting timeline, and the fixed gas pricing detail since that is the point most likely to matter to someone evaluating Redbelly from an institutional angle. It ends with the same disclaimer as the full report: This is not a price prediction.

Turning four files into something reviewable

Once all four parts were done, I had a pile of files: a PDF and DOCX report, a nine-file DeFiLlama submission kit, a five-file RWA.xyz kit, and a markdown explainer. Technically complete, but not something a reviewer could actually navigate without opening a dozen separate downloads in the wrong order.

The fix was structural rather than cosmetic. Everything went into a single GitHub repository, organized into four numbered folders under one parent directory, with a top-level README that summarizes each part in one line and links straight to it. One link on the task board, and a reviewer can walk through all four deliverables in the order they were built, with the reasoning for each visible right alongside the work.

From there, I built a companion website that renders the entire submission inline rather than forcing anyone to dig through a file tree. The tokenomics report displays as an embedded, scrollable PDF directly in the page. The DeFiLlama and RWA.xyz kits render through a tabbed file browser with syntax highlighting, so the actual code and documentation are readable without downloading anything first. The explainer sits in full on the page itself, since it is short enough not to need collapsing.

The design intentionally avoids the generic crypto dashboard look. It reads more like an audited case file: ink navy background, paper-white content cards, and a red verification stamp mark used sparingly only next to claims that are actually backed by a checked source.

What the task actually taught

The brief for TASK-11 read like four independent chores. In practice, the real work was verification, not production.

The DeFiLlama bug would never have surfaced without actually running the manual check instead of trusting that the adapter worked because it compiled. The RWA.xyz finding would have been a wasted submission kit for a listing that unlike DeFiLlama turned out to still be genuinely missing, if I had not checked the current state of both platforms before writing anything. And the naming collision between rwa.io and RWA.xyz would have been an easy, confusing mistake for a reviewer to make without a flag pointing it out directly.

None of that shows up in a task description. It only shows up if you go check.

Repository: github.com/0xDarkSeidBull/dao-redbelly/tree/main/task11-rbnt-report

Live site: daotask11.test-hub.xyz

All tokenomics figures in this report are sourced from the official Redbelly Network whitepaper ("Redbelly Network: An accountable data layer for AI powered financial services," 28 August 2025) unless otherwise footnoted. This report contains no price predictions or investment advice.