USG — Risk Report
Category: Stablecoin | Peg Mechanism: Hybrid (CDP + PegKeepers) | Issuer: Tangent Finance
Live data: USG Backing Dashboard — hourly per-market collateral, liquidation headroom, oracle divergence, and PegKeeper state.
USG is a crvUSD fork issued by Tangent Finance on Ethereum. Users mint it against collateral in isolated lending markets, and a pair of PegKeepers defends the dollar in Curve pools. Every unit is collateralised on-chain, there is no off-chain reserve, and no attestation is required to verify it — the entire book can be read from contracts. That is the strongest thing about this asset, and it is why this report can be specific where reports on custodial stablecoins cannot.
The headline is healthy and the distribution is not. At 2026-08-27 the aggregate collateral ratio is 131.56% conservative / 155.48% inclusive, with zero bad debt in any of the sixteen markets. Read alone, that describes a comfortable book.
Score breakdown
| axis | score | reasoning |
|---|---|---|
| Stability | 5.0 | CDP + PegKeeper is proven in crvUSD, but keeper capacity is $0 and pools are 73.6% skewed |
| Backing | 5.0 | Fully on-chain and verifiable, zero bad debt — but 74.9% of the book is within 10% of liquidation |
| Liquidity & Exit | 3.5 | One venue, no quoted depth, about $1.0M counter-side inventory against $4.34M supply |
| Dependencies | 4.0 | Debt-weighted lookthrough across the 12 live markets, each LP taken at its worse leg; no point value is quoted, since the per-market debt and the constituent scores are from different reads. Scored below that on purpose: every position is 3–4 layers deep across 12+ external protocols, and MIN(coin price) removes the diversification a pair would otherwise give. Weakest named constituent is Resupply reUSD at 3.5 — 9.1% of debt, and in two of the four thinnest markets |
| Contract & Admin | 3.5 | Asset permissioning, parameter setting and governance all terminate at one 3-of-5 Safe with no execution delay; repointing collatOracle moves every position’s health in one transaction, against a book where 74.9% sits within 10% of liquidation. The only 3-day delay covers replacing the PegKeeper admin, not the risk setters |
| Issuer | 4.0 | Transparent data, unreviewed parameters, young sub-$5M protocol |
| Overall | 4.0 | Unchanged. Axis 5 makes an authority surface explicit that was previously priced inside the Issuer row — it is a re-scoping of where the fact sits, not a new adverse read of the asset |
1 · Stability — 5.0
USG trades at $0.99534, a −0.466% discount. The mechanism meant to close that is the PegKeeper pair, and its deployable defence capacity is currently $0 — both keepers report debt() = 0, meaning no POL-minted USG is outstanding to burn.
Meanwhile the Curve pools they operate in are 73.56% USG on the worst side, against 50% at balance. A pool that heavy on one asset is one where sellers have already moved through, and the counter-side inventory — $1,038,555 of USDC and frxUSD — is LP-owned, not protocol-deployable. It is exit liquidity for USG holders, which matters, but it is not peg defence and should not be counted as such.
No executable depth ladder is quoted for USG at any venue. Pool TVL is $3.9M, but TVL is not depth, and this report does not treat it as such. The liquidity score of 3.5 reflects observed exit capacity — a single-venue, skewed pool with roughly $1.0M of counter-side inventory — not the headline TVL.
2 · Backing — 5.0
Recomputing the aggregate from the per-market data returns 131.56% — it is the debt-weighted mean of the individual markets, which is exactly why it conceals their spread.
| market | debt | CR | headroom to liquidation |
|---|---|---|---|
| StakeDao — reUSD/sDOLA | $248,410 | 119.5% | 1.55% |
| StakeDao — frxUSD/OUSD | $65,542 | 114.9% | 1.95% |
| StakeDao — frxUSD/sDOLA | $351,293 | 120.0% | 2.27% |
| Convex FXN — reUSD/fxUSD | $230,947 | 121.0% | 2.73% |
| Curve — RLUSD/USDC | $473,015 | 114.7% | 4.68% |
| StakeDao — BOLD/USDC | $499,496 | 115.4% | 5.05% |
| Convex FXN — fxUSD/USDC | $699,999 | 116.1% | 5.64% |
| StakeDao — USDT/crvUSD | $181,373 | 127.1% | 7.72% |
| StakeDao — frxUSD/sUSDS | $500,690 | 120.3% | 8.89% |
| StakeDao — reUSD/scrvUSD | $247,806 | 144.1% | 18.37% |
| Curve — PYUSD/USDC | $500,490 | 134.6% | 18.82% |
| StakeDao — cbBTC/WBTC | $65,386 | 180.5% | 33.85% |
| StakeDao — msETH/WETH | $276,112 | 261.6% | 49.87% |
$3,250,770 — 74.9% of the $4.34M book — sits within 10% of its liquidation threshold. $1,369,209, or 31.5%, sits within 5%.
The two markets at the bottom of the table are the reason the average clears 130%. msETH/WETH and cbBTC/WBTC together are $341,498 — 7.9% of debt. Strip them and the remaining 92% of the book averages 121.74%, against a debt-weighted liquidation floor of 113.15%. A single aggregate ratio of 131.56% describes a portfolio in which almost no individual position is that comfortable.
One qualification that runs against the borrower: these are market-wide aggregates, verified against the contracts — totalCollateral() and totalDebtShares() reproduce the published figures exactly. Each row is the distance to that market’s average LTV reaching its threshold. Individual borrowers above the market average are closer, and some may already be liquidatable. The table understates how soon the first liquidation lands, not how late.
Supply reads ten times larger than it is
totalSupply() returns $44,340,302. Real circulating supply is $4,340,566. The difference is the pre-minted PegKeeper ceiling buffer — a crvUSD inheritance in which the contract mints its own ceiling up front and holds it. Any tool that reads totalSupply() on a crvUSD fork and divides will understate the collateral ratio by an order of magnitude. Real supply equals CDP debt exactly; with POL deployment at zero, 100% of circulating USG is CDP-backed. Staked USG (sUSG) holds $1,064,103, a subset of circulating supply rather than an addition to it.
3 · Liquidity & Exit — 3.5
USG trades on two Curve PegKeeper pools and nowhere else, and the exit is now measured rather than described. Quoted 2026-09-11 with get_dy on each pool — the call a seller’s transaction actually makes — against a $4.46M supply:
| selling | USG → USDC | USG → frxUSD | split across both |
|---|---|---|---|
| $10,000 | −51.5 bps | −45.1 bps | — |
| $100,000 | −64.8 bps | −54.0 bps | — |
| $250,000 | −108.0 bps | −77.7 bps | — |
| $300,000 | −136.9 bps | −90.3 bps | −67.7 bps |
| $400,000 | −276.1 bps | −131.3 bps | −78.4 bps |
| $500,000 | −929.0 bps | −229.7 bps | −92.9 bps |
| $600,000 | — | — | −113.6 bps |
| $1,000,000 | −5,242 bps | −3,930 bps | — |
⚠️ The exit is expensive at every size, and that is the finding a depth headline would hide. Selling $10,000 costs about half a percent. This is not a deep-at-small, thin-at-size profile — the floor cost is high and the cliff is steep on top of it.
✅ But a seller who splits across both pools does far better than either alone, and it is worth stating because the obvious method understates it. Each pool crosses 2% on its own between $350K and $400K (USDC) and between $450K and $500K (frxUSD). Split, $600,000 clears inside 1.2% — about 13.5% of supply. A single-pool ladder would read roughly 2.5× worse than the executable exit, because the pools are independent and a seller uses both.
⚠️ Two limits on these figures, stated rather than buried. They are a single-block snapshot, and PegKeeper pool balances move with keeper activity — this is a dated reading, not a standing property. And the split column quotes an even split, not an optimal one, so it is a floor on best execution rather than the best a solver would find.
What backs the exit is inventory, not depth-on-paper: $471,641 of USDC and $611,245 of frxUSD sit opposite USG across the two pools — $1,082,886 a seller can execute into, against $4.46M outstanding. ⚠️ The USG side of those same pools is not exit liquidity for a seller, and USG is 71–73% of each pool, so the counter-side is the smaller half of a small venue.
✅ Nothing here is broken and there is no evidence of a failed exit. The measurement says what an exit costs, not that one has gone wrong.
4 · Dependencies — 4.0
Scoring the upstream exposure is a lookthrough rather than a judgement: across the 12 live markets, debt-weighted, with each collateral LP taken at its worse leg. ⚠️ No single point value is quoted for the result — the per-market debt figures and the constituent scores come from different reads, and the book has moved materially between them. Sixteen of the eighteen constituents already carry their own assessments; the weakest named one is Resupply reUSD at 3.5, which is 9.1% of debt and sits in two of the four thinnest markets.
⚠️ The worse-leg rule is not a convention imposed from outside — it is USG’s own oracle arithmetic. The market oracle is virtual_price × MIN(coin price), so the weaker constituent of a pair sets the mark, and the mark is what sets liquidation. You take the worse leg’s risk without the benefit of holding two assets, which is the opposite of what holding an LP position normally buys you.
The axis scores 4.0, below the 5.44 lookthrough, and the gap is deliberate: constituent quality is not the binding constraint here, structure is. Every position is 3–4 layers deep — USG, then a StakeDAO/Convex/Curve wrapper, then a Curve pool, then two stablecoins, then their own backing — across 12+ external protocols. Set that against the distribution above, where 74.9% of the book sits within 10% of liquidation, and the depth of the stack is what decides how far a single upstream problem travels.
reUSD is the concentration, and there are two tokens with that ticker
Three markets are collateralised by pools containing reUSD — reUSD/sDOLA, reUSD/scrvUSD, and reUSD/fxUSD, totalling $727,162, or 16.8% of the book. Two of them are among the four thinnest positions USG holds, and the single thinnest market in the entire book, at 1.55% headroom, is reUSD/sDOLA.
The protocol’s own monitoring flags this: “reUSD depeg −1.02% — USG collateral cascade leading indicator.” That is the correct reading. USG’s collateral is not the stablecoins themselves but Curve LP tokens containing them, and an LP position holding a depegging asset does not lose half its value — arbitrageurs sell the broken leg into the pool, so the pool accumulates it and the loss approaches the full position.
The reUSD in question is Resupply’s, at 0x57aB1E0003F623289CD798B1824Be09a793e4Bec. It is not Re Protocol’s reUSD, a structurally unrelated NAV-accruing reinsurance token that shares the ticker. Anyone assessing this exposure by symbol will reach the wrong asset. Resupply’s reUSD is itself a CDP stablecoin, so USG’s collateral here is another protocol’s debt token — a second layer of the same risk shape.
Aggregating all thirteen active markets by underlying asset, USDC touches 50.1% of the book, fxUSD 21.4%, frxUSD 21.1%, reUSD 16.8%, sDOLA 13.8%. Seventeen distinct assets in total. These sum past 100% because each market has two legs and either can impair it.
Oracles and paused markets
USG prices collateral through per-market oracles, and the dashboard independently recomputes each pool’s NAV from its underlying balances to compare against them. At this reading no market exceeds the 1% divergence threshold: zero divergent markets, with a maximum observed overvaluation of 0.57%.
That is a genuine improvement rather than a quiet one — earlier the same day, two markets were divergent with a maximum overvaluation of 1.11%, the largest in StakeDao’s msETH/WETH pool. Divergence at that scale is not alarming in a market carrying 49.87% headroom. It matters because the same oracle design serves markets with under 2% headroom, where a 1% overvaluation is most of the remaining buffer. The check is worth watching precisely when it reads clean, because that is when the thin markets are being priced on the same basis as the comfortable ones.
Two of the sixteen markets are paused. Tangent’s paused-settings semantics let existing positions persist while new borrowing stops, so a pause is a freeze rather than a wind-down — the collateral stays, the exposure stays, and the market can be reopened. Neither paused market carries debt today.
5 · Contract & Admin — 3.5
The oracle question a reader usually asks is where does the price come from. On USG the sharper question is who can change which feed a market reads — the parameter layer rather than the price feed. A sound oracle behind a swappable pointer is not a sound price.
Three distinct powers all terminate at the same 3-of-5 Safe, with no execution delay between a decision and its effect: asset permissioning, parameter setting, and governance. Walked at block 25915396, setMaxLTV, setLiquidationThreshold and setCollatOracle on a live market succeed when simulated from the Safe and revert OnlyOwner from a null address — which establishes who is able to call them, not that anyone has.
⚠️ The consequence is specific to this book rather than generic. The market oracle is virtual_price × MIN(coin price), so repointing collatOracle moves every position’s health in a single transaction. Set that against the distribution above: 74.9% of the book sits within 10% of its liquidation threshold and the thinnest markets have 1.55% to 2.27% of headroom. A parameter change that would be an inconvenience in a comfortably collateralised book is, here, the same size as the buffer.
A delay does exist, and it protects the wrong thing. The PegKeepers carry a 3-day ADMIN_ACTIONS_DELAY — but it covers replacing the PegKeeper admin, not the risk setters. The powers that move liquidation thresholds and oracle assignment have no delay at all.
None of this is evidence of misuse, and every part of it is readable on-chain by anyone. It is the standing surface: a fully verifiable book whose risk parameters are one multisig transaction away from changing, with no notice window in between. For an asset whose whole strength is that the collateral can be read from contracts, this is the part that cannot be.
6 · Issuer — 4.0
Tangent Finance is a small, pseudonymous-adjacent team with no published reserve attestation — none is needed, since backing is fully on-chain, but there is also no third-party review of the market parameters that decide when liquidations trigger. The issuer score of 4.0 reflects a young, sub-$5M protocol whose data is unusually transparent and whose oversight is not: no external party checks the parameters, and no published process governs who sets them. The authority those parameters sit behind is scored separately, on Contract & Admin above — the two are different questions, and pricing the multisig on both would count it twice.
How this compares to its parent
USG inherits crvUSD’s architecture, and the differences are mostly ones of scale and collateral quality. crvUSD’s collateral is majority blue-chip — ETH, WBTC, staked ETH derivatives — priced by an AMM designed to liquidate continuously rather than at a cliff. USG’s collateral is almost entirely other protocols’ stablecoin LP tokens. That trades one risk for another: USG’s book is far less exposed to crypto beta, and much more exposed to a single peer stablecoin breaking.
The practical consequence is that USG’s failure mode is not a market crash. It is one upstream issuer having a bad week. A 20% ETH drawdown would barely touch this book; a 3% reUSD depeg would put three markets at their thresholds at once.
The second inherited difference is defensive capacity. crvUSD’s PegKeepers currently carry substantial debt and therefore substantial burn capacity. USG’s carry none, which means the peg is being held by market-making and redemption arbitrage rather than by the mechanism the design nominates for it. At a −0.47% discount that is working. It is untested at a wider one.
What holding USG involves
USG is small — $4.34M of real supply — and concentrated in a single Curve venue. That combination sets the practical position size well below what the collateral ratio alone would suggest. A holder large enough to matter to the pool is a holder who cannot exit it quickly, and at 73.56% USG on the worst side, some of that exit has already happened.
The staking wrapper, sUSG, holds $1,064,103 — about 24.5% of supply. Staking does not change the underlying collateral risk; it adds a wrapper layer and a yield source on top of the same book.
What would change this view
Upward: liquidation headroom widening across the thin markets rather than through new comfortable ones; PegKeeper debt becoming non-zero, giving real burn capacity; a second liquidity venue; independent review of market parameters.
Downward: a reUSD depeg deepening past ~2%, which would put the three reUSD markets at their thresholds simultaneously; pool skew above ~85%; any market entering bad debt, which would show that liquidations are not clearing at these headroom levels.
The thing to watch is not the collateral ratio. It is the gap between the aggregate and the thinnest markets. That gap has been widening while the headline stayed flat, and the headline is where most monitoring stops.
Figures measured 2026-08-27 from on-chain contract reads. Zero bad debt confirmed across all sixteen markets. Live figures on the dashboard linked above.