RWA Infrastructure11 min read
MB
Editorial Team
·November 17, 2026

Can a Proof of Reserve Feed Stop a Token From Over-Minting?

Chainlink's SecureMintPolicy is a mint gate: before each mint it reads a Proof of Reserve feed and rejects the transaction if current supply plus the new amount would exceed the margin-adjusted reserve. The cap is only as strong as four choices the issuer makes: the staleness threshold, the margin mode, the feed and its data source, and who holds the keys that can change them.

TL;DR — Key Takeaways

  • ✓The check: SecureMintPolicy reads latestRoundData() from a reserve feed and rejects a mint if totalSupply plus the amount exceeds the margin-adjusted reserve. It also rejects negative reserves and, if configured, stale data.
  • ✓Staleness: A threshold of 0 disables the age check, so reserve data of any age is accepted. Chainlink's docs suggest a value that matches or slightly exceeds the feed's heartbeat.
  • ✓Margin: Five modes. With reserves of 1,000 the limit ranges from 900 (10% positive) to 1,100 (10% negative). Negative modes let supply exceed reported reserves by design.
  • ✓Feeds: One feed per policy instance. Separate instances each compare the whole supply to their own feed, so on our reading of the code they do not add reserves across custodians.
  • ✓Limits: The gate never burns, pauses or redeems when reserves fall, the policy owner can change every parameter, and a self-reported feed carries Chainlink's own accuracy disclaimer.

Ready to get started?

Join others who are already using our platform.

Can a Proof of Reserve Feed Stop a Token From Over-Minting?

What does Chainlink's SecureMintPolicy check before a mint?

It runs four checks in order: the reserve value is not negative, the data is not older than the staleness threshold, the reserve is scaled to the token's decimals, and totalSupply plus the mint amount does not exceed the margin-adjusted reserve.

A Proof of Reserve (PoR) feed is an on-chain data feed that reports the reserves behind a token. On its own it only publishes a number. SecureMintPolicy, part of the Chainlink ACE policy library, turns that number into a hard limit at issuance. This page reads the contract at commit 60f04505 of the chainlink-ace repository (23 July 2026, typeAndVersion “SecureMintPolicy 1.2.0”) alongside the SecureMintPolicy reference page. Chainlink's documentation describes ACE as a private beta.

Step in run()Code (SecureMintPolicy.sol)Rejection message
Read the feedL403: latestRoundData() returns the reserve and updatedAtNone at this step
Reject a negative reserveL406-407“reserve value is negative”
Reject stale dataL410-411, only if maxStalenessSeconds is above 0“reserve data is stale”
Scale feed decimals to token decimalsL414-426None at this step
Compare supply to the limitL427-428: amount + totalSupply() against the margin-adjusted reserve“mint would exceed available reserves”

If every check passes, the policy returns Continue, not Allowed. It never approves a mint by itself. It can only stop one.

How should the staleness threshold be set?

Chainlink's documentation says it should match or slightly exceed the feed's heartbeat, the longest interval between updates. A value of 0 switches the age check off, so a feed that stopped updating still sets the mint cap.

The code at L410 reads maxStalenessSeconds > 0 && block.timestamp - updatedAt > maxStalenessSeconds. Both conditions must be true to reject. The parameter is set in configure() at deployment and can be changed later through setMaxStalenessSeconds(). Neither path enforces a minimum, so 0 is a valid setting. The heartbeat of each PoR feed is listed on Chainlink's Data Feeds pages, and this page does not quote any specific feed's value.

The trade-off runs both ways. A threshold far above the heartbeat lets old reserve data authorise new supply. A threshold below the heartbeat makes mints fail between normal updates, which looks like an outage to the operations team. Stale reserve data is one form of the feed-governance failures described in the analysis of 2026's oracle configuration failures, where the code worked and the settings did not.

“Your staleness threshold should typically match or slightly exceed the heartbeat. Setting the staleness threshold to 0 disables the check entirely — the policy will accept reserve data of any age.”

Chainlink documentation, SecureMintPolicy, Staleness threshold

What do the five reserve margin modes allow?

With reserves of 1,000, the five modes set the mint limit anywhere from 900 to 1,100 in Chainlink's own examples. Positive modes keep a buffer below reserves; negative modes let supply exceed reported reserves.

ModeFormula (L355-374)Docs example, reserves = 1,000
Nonelimit = reserves1,000
PositivePercentagereserves × (10,000 − margin) / 10,00010% gives 900
PositiveAbsolutereserves − margin, or 0 if reserves are below the margin50 gives 950
NegativePercentagereserves × (10,000 + margin) / 10,00010% gives 1,100
NegativeAbsolutereserves + margin50 gives 1,050

Percentage amounts are in basis points: 1234 means 12.34% and the maximum is 10,000. Absolute amounts must be above 0. At the maximum, NegativePercentage allows supply of twice the reported reserve. The documentation frames negative margins as “operational flexibility”, for example to cover the gap between a mint and the next feed update. An investor reading “backed 1:1” would not expect that headroom unless it is disclosed.

A comment at L360 warns that PositivePercentage “may round to zero for very small reserves with high margins”, with the example of reserves of 1 and a margin of 9,999. The setter's comment recommends absolute modes for low-precision feeds.

Can one policy cover reserves held at several custodians?

No. Each SecureMintPolicy instance reads one feed. Chainlink's documentation says to deploy one instance per reserve source and add them all to the same policy chain, which means every instance must pass.

“Each SecureMintPolicy instance supports one feed. If your token is backed by multiple reserve sources, deploy a separate SecureMintPolicy instance for each feed and add them all to the same policy chain.”

Chainlink documentation, SecureMintPolicy, Proof of Reserve feed

Our reading of the code adds a consequence the documentation does not spell out. Each instance compares the token's entire totalSupply() to its own feed. It does not know the other feeds exist. Take a token with 900 units outstanding and reserves split between custodian A (600) and custodian B (400). Instance A sees 900 against 600 and rejects. Instance B sees 900 against 400 and rejects. The chain never compares supply to the combined 1,000. In practice the effective cap becomes the smallest single feed, so the gate fails closed rather than open.

Two designs avoid this: one feed that reports aggregated reserves across custodians, or a custom policy that sums several feeds. The policy engine at the same commit caps each protected function at 8 policies (MAX_POLICIES in PolicyEngine.sol), which also limits how many per-feed instances one mint function can carry. This is from reading the code; no fork test was run.

Why do decimals and units decide whether the comparison means anything?

The policy compares two integers. If the token decimals are wrong, or the feed counts something other than token-equivalent units, the comparison runs without error and gives the wrong answer.

The contract's header (L29-32) still says the feed “MUST report values in the same decimals as the protected token.” The run() function then scales between feed decimals and token decimals (L414-426), and the documentation says the policy “scales between the feed's decimals and the token's decimals automatically.” The code is what executes, so the scaling applies; the header comment is out of date. The real risk sits elsewhere. Token decimals are set by the issuer, between 1 and 18. If the token exposes decimals(), the policy checks the value against it. If not, the value is accepted as given, and the documentation warns that a wrong value can mean “potentially allowing over-minting or blocking valid mints.” The same unit problem on the subscription side is covered in the guide to token decimals and RWA subscription amounts.

Units are a design check, not a code check. Chainlink says PoR feeds report “in units of measurement such as ounces (oz) or number of tokens.” A gold token minted at 1 token per ounce can be compared to an ounce feed. A fund share token priced at a NAV of, say, $1.07 cannot be compared to a feed of dollar reserves without converting one side. With $1,070 of reserves backing 1,000 shares, a dollar feed would let the gate allow 1,070 shares, 70 more than the reserves back. At a NAV below $1, the same mistake blocks valid mints. The feed must report reserves in share-equivalent units, or the policy must be fed a value that has already been converted.

Which Proof of Reserve feed types carry Chainlink's self-reported warning?

Three of the five types Chainlink lists: self-reported off-chain reserves read from an issuer-hosted API, and both cross-chain types where the issuer declares which wallet addresses it owns.

Feed type (SmartData docs)Where the number comes fromChainlink disclaimer
Off-chain, third-partyAn auditor, accounting firm or other third party verifies reservesNo self-reported warning
Off-chain, custodianPulled directly from the bank or custodianNo self-reported warning
Off-chain, self-reportedAn API the token issuer hosts“Additional risks”; Chainlink Labs not responsible for accuracy
Cross-chain, wallet address managerIssuer self-reports its addresses in an IPoRAddressList contractSame warning
Cross-chain, wallet addressIssuer reports its addresses through a self-hosted APISame warning

A mint gate fed by a self-reported source checks the issuer against the issuer. It still stops a mint that exceeds what the issuer itself reports, which catches operational errors and compromised minter keys. It does not prove the reserves exist. The SmartData page adds a separate single-source warning: many PoR feeds rely on one data provider, such as an issuer, custodian or auditor.

“Reserve data reported by an asset issuer's self-hosted API carries additional risks. Chainlink Labs is not responsible for the accuracy of self-reported reserves data. Users must do their own risk assessment for asset issuer risk.”

Chainlink documentation, SmartData, Types of Proof of Reserve Feeds

What does the mint gate not do?

It does not act when reserves fall, it does not run on functions it was not attached to, and it can be skipped by an earlier policy in the chain. Each limit follows from the code at commit 60f04505.

  • No action after a drop. The policy runs only when a protected call is made. If the feed falls below outstanding supply, nothing is burned, paused or redeemed. The next mint is rejected; existing tokens stay where they are.
  • Only the selectors it is added to. The engine stores policies per contract and per function selector. A token with a second mint path, such as a batch mint or a bridge mint, needs the policy on each selector.
  • Order matters. The engine runs policies in sequence and stops at the first one that returns Allowed. SecureMintPolicy only returns Continue. The library's BypassPolicy returns Allowed when every checked address is on its list, so a BypassPolicy placed earlier on the mint selector would skip the reserve check for those addresses.
  • A default still applies. If every policy returns Continue, the engine falls back to a default allow or reject set by its admin. With default reject, a chain that contains only SecureMintPolicy blocks every mint.

Every parameter can also change after launch. setReservesFeed, setReserveMargin, setMaxStalenessSeconds and setTokenMetadata are onlyOwner. The policy engine's ADMIN_ROLE can remove the policy from a selector, upgrade its implementation and transfer its ownership. A gate that the minter's own organisation can loosen in one transaction protects against errors more than against the issuer.

What should an issuer disclose about a reserve mint gate?

Six items, each readable from chain state: the feed and its source type, the staleness threshold, the margin, the decimals setting, the protected mint functions and policy order, and the keys that can change them.

ItemHow an auditor reads itWhat weakens the gate
Reserve feed and source typereservesFeed(), then the feed's listing in Chainlink's SmartData directorySelf-reported source; single provider
Staleness thresholdmaxStalenessSeconds() compared with the feed heartbeat0, or far above the heartbeat
MarginreserveMarginMode() and reserveMarginAmount()Any negative mode
Token decimalsTokenMetadataSet event versus the token contractToken without decimals(), so no on-chain cross-check
Protected selectors and orderThe engine's policy list for each mint selectorUnprotected mint path; a BypassPolicy ahead of the gate
Change authorityPolicy owner() and engine ADMIN_ROLE holdersSingle key; no timelock

A reserve gate answers how many tokens may exist. Identity and transfer rules answer who may hold them. A disclosure pack should list both sets of parameters with their owners, alongside the periodic reports covered in the RWA investor reporting and disclosure requirements guide. An auditor who can read these values from chain state can tell in minutes whether “backed 1:1, verified on-chain” describes an enforced limit or a dashboard.

Who is this for, and what was not checked?

It is for issuers of asset-backed and RWA tokens deciding whether to gate minting on a reserve feed, and for the auditors who review that setup. Everything above comes from reading Chainlink's documentation and the source at one commit; no fork test was run.

We did not check which live tokens use SecureMintPolicy, and this page does not claim any do. We did not read the older Secure Mint pattern that checks reserves inside the token contract. The library is labelled beta, so parameters and behaviour may change after commit 60f04505. The multi-feed and policy-order points are inferences from the code, not statements from Chainlink. This is technical analysis, not legal or audit advice.

Frequently Asked Questions

What does Chainlink's SecureMintPolicy do?

It is a policy in the Chainlink ACE policy library that runs before a mint. It reads the latest value from a Proof of Reserve feed through latestRoundData(), scales it to the token's decimals, applies a configured margin and rejects the mint with "mint would exceed available reserves" if totalSupply plus the mint amount is above that limit. It also rejects negative reserve values and, if a threshold is set, stale data.

What happens if the SecureMintPolicy staleness threshold is set to 0?

The staleness check is skipped. The code only compares block.timestamp minus updatedAt with maxStalenessSeconds when maxStalenessSeconds is above 0, and Chainlink's documentation says a value of 0 means the policy will accept reserve data of any age. A feed that stopped updating months ago would still be used as the mint cap. The documentation suggests matching or slightly exceeding the feed's heartbeat.

Can supply exceed reserves under a Chainlink Secure Mint policy?

Yes, if the issuer chooses a negative margin. NegativePercentage lets supply exceed reserves by up to 100% (10,000 basis points), and NegativeAbsolute by any fixed amount above zero. With reserves of 1,000, a 10% negative margin sets the limit at 1,100 and a 50-unit negative absolute margin sets it at 1,050. The margin mode and amount should be disclosed, because the default reading of the gate is 1:1.

Does SecureMintPolicy burn tokens or pause the token if reserves fall?

No. It runs only when a protected mint function is called, and its only action is to reject that call. It has no hook that runs when the feed value drops, so if reserves fall below supply, existing tokens stay outstanding. Redemptions, burns or a pause need separate controls. The gate also checks nothing on transfers, so it says nothing about who may hold the token.

Is a Chainlink Proof of Reserve feed independent of the issuer?

Not always. Chainlink's SmartData documentation lists third-party, custodian and self-reported sources for off-chain reserves, and wallet address manager or wallet address lists for cross-chain reserves. For self-reported types it states that Chainlink Labs is not responsible for the accuracy of self-reported reserves data. It also says many Proof of Reserve feeds rely on a single data provider.

Ready to get started?

Join others who are already using our platform.