Vocabulary
FACTA answers with these terms and no others. Each states what was observed and the test or threshold behind it, and none is advice. A term is published with its test before any answer carries it.
Agent liveness
- Method
| Term | Definition | Test |
|---|---|---|
Live service | The registration file is valid, and at least one declared endpoint answered correctly: an MCP handshake, an agent card that parses, or a payment-required response with valid terms. | Set by the job register.liveness. An MCP endpoint is asked in both generations of the protocol, and either answer counts. Valid payment terms name at least one way to pay with a scheme, a network, an asset, an amount and a recipient, all well formed. At most 4 declared services are probed per identity, with at most 2 requests each. |
Site only | The registration file is valid, at least one declared endpoint answered over HTTP, and none answered as the service it claims to be. | Set by the job register.liveness. Answered over HTTP means a response with status 200 to 399, 401, 402, 403 or 405. A 404, a 410 and every 5xx are no answer. A payment-required response without valid terms counts as answered over HTTP. |
Card only | The registration file is valid, and it declares no service that could be asked, or none of those asked answered at all. | Set by the job register.file-fetch when the file declares no service that could be asked, otherwise by register.liveness. A term never gets worse on one failed attempt: a 429, a 5xx, a timeout or a network error is tried again six hours later, and only the third failure in a row lowers the term. |
No card | The identity has no registration URI. | Set by the job register.crawl, which reads tokenURI for every identity once in each pass of its full read. The URI it returns is empty. |
Unreachable | The identity has a registration URI, and the file could not be read under the fetch rules in three attempts, or what came back is not a registration file. | Set by the job register.file-fetch. A 429, a 5xx, a timeout or a network error is one attempt, and the next comes six hours later. A 404, a 410, a URL the fetch rules refuse and a body that is not a registration file are results at once. A host whose operator asked the crawler to stay away is not fetched, and its identities carry this term with that reason. |
Token listing
- Method
| Term | Definition | Test |
|---|---|---|
Issuer-listed | The address is in Robinhood's Stock Token registry or in Ondo's token file as of the last successful read of that list. No other issuer's list is read. | Rule 2 of 4, decided in order. Both lists are read every 6 hours. When the last successful sync of a list is older than 12 hours, the answer also names that list as stale. |
Not issuer-listed | The address is in neither Robinhood's nor Ondo's list, and its symbol has the same comparison key as the symbol of a token they list, or its name follows the naming pattern of one of them for a stock it lists. No other issuer's list is read. | Rule 3 of 4, decided in order. The comparison key normalises the text, folds look-alike letters of other scripts to Latin, keeps only letters, digits and the dot, and upper-cases. The answer says whether the match is on the symbol or on the name, and gives the listed address. When the list of the issuer it matches is past its freshness limit or awaits review, the term is Unknown instead. |
Not a stock token | The address answers as a token, is in neither Robinhood's nor Ondo's list, and neither its symbol nor its name matches a stock they list. No other issuer's list is read. | Rule 4 of 4, decided in order: every token the first three rules did not decide. A token that another issuer lists is in neither list, so it gets this term until the method reads that issuer's list. |
Unknown | The address has no code, or does not answer symbol(), name() and decimals() as a token. The term also applies when the address matches a token Robinhood or Ondo lists, by its symbol or by its name, while that issuer's list is past its freshness limit or awaits review. | Rule 1 of 4, decided in order, and the exception to rule 3: an old copy of a list cannot show that an address is missing from it. In that second case the answer names the list as stale. |
Market state
- Method
| Term | Definition | Test |
|---|---|---|
Open | The token has a price feed, its oracle is not paused, the session schedule is running, and the feed has not been quiet for longer than its heartbeat plus a margin of 600 seconds, counted in session time. | Rule 4 of 4, decided in order. A quiet feed alone does not change the term: every equity feed's heartbeat is 86,400 seconds, and only a feed quiet for longer than its heartbeat plus a margin of 600 seconds, counted in session time, answers Closed. The margin is there because a heartbeat update lands a little after the heartbeat. When the feed read or the oraclePaused() read fails there is no term, and the answer names the source as stale. |
Closed | The token has a price feed, its oracle is not paused, and the session schedule is not running. The term also applies when the schedule is running and the feed has been quiet for longer than its heartbeat plus a margin of 600 seconds, counted in session time. | Rule 3 of 4, decided in order, and the exception to rule 4. Session time is counted from the later of the feed's last update and the start of the current session. The answer names the signal that decided. The session schedule and its holidays are a reviewed data file. |
Paused | The token has a price feed, and the token contract reports its oracle as paused. | Rule 2 of 4, decided in order: oraclePaused() is true. The issuer's price endpoint also reports trading halts, and version 0.1.0 does not read it. |
No feed | No price feed is published for the token. | Rule 1 of 4, decided in order, and the common case: on 2026-11-10 there were 35 feeds for 218 listed tokens. A token with no feed keeps this term when its oracle is paused, and the answer then says so in a note. |
Pool Check
- Method
| Term | Definition | Test |
|---|---|---|
Organic | Over the window, round-trip trading is under 15% of the pool's reported volume and turnover is under 10 times its liquidity a day. | Default thresholds, version 0.1, to be calibrated on the first 100 pool-days. Round-trip volume is, for each trader, twice the smaller of its buy and sell volume in the window, counted when its net change is under 5% of its gross. Turnover is reported volume divided by the average value of in-range liquidity. |
Mixed | The pool has at least 20 swaps in the window and meets neither the thresholds for Organic nor those for Recycled. | Default thresholds, version 0.1: everything between the two other tests. |
Recycled | Over the window, round-trip trading is above 50% of the pool's reported volume, or turnover is above 50 times its liquidity a day, or one swap is larger than the pool's liquidity. | Default thresholds, version 0.1, to be calibrated on the first 100 pool-days. The term reports what the swaps of the window measured against these thresholds. It names no trader and no intent. |
Insufficient data | The pool has fewer than 20 swaps in the window, so none of the other three terms is computed. | Default threshold, version 0.1: 20 swaps. It is decided before the other three. |
Income source
- Method
| Term | Definition | Test |
|---|---|---|
Third-party | The payer meets none of the tests for Financing, Related-party or Single-purpose. | It is what remains after the three other tests, each with its default threshold of version 0.1. |
Related-party | The payer is another agent wallet of the agent's owner, or its first funder is the owner or the agent wallet, or it shares a first funder with them. The term also applies when the payer received at least the paid amount from the owner or the agent wallet in the 30 days before it paid. | Default threshold, version 0.1: 30 days. A first funder that sent value to more than 500 addresses, or is a labelled bridge or exchange, is ignored in the shared-funder test. A transfer from the owner itself is not counted as income at all, as the income-statement draft ERC-8458 counts. |
Single-purpose | The payer was first funded within 7 days before its first payment to the agent and has never paid anyone else. | Default threshold, version 0.1: 7 days. |
Financing | The sender is a known lending pool. Loan proceeds are not counted as income. | It has no threshold: the test is whether the sender is a lending pool the method knows. |
Audit opinion
- Method
| Term | Definition | Test |
|---|---|---|
Unqualified | Over the audit window, none of the thresholds for Disclaimer, Adverse or Qualified was met. | Default window: 90 days. Code computes the opinion from the figures of the audit. No model chooses it. |
Qualified | Over the audit window, Third-party income is from 40% to 70% of gross income, or the top payer accounts for more than 60% of Third-party income, or more than 50% of loans were repaid within 24 hours without the borrowed funds leaving the wallet. | Default thresholds, version 0.1, to be calibrated on the first 50 audits. It applies when the thresholds for Disclaimer and for Adverse were not met. |
Adverse | Over the audit window, Third-party income is under 40% of gross income, or a loan defaulted in the window, or funded round trips are above 25% of gross income. | Default thresholds, version 0.1, to be calibrated on the first 50 audits. It applies when the thresholds for Disclaimer were not met. |
Disclaimer | The audit window does not support an opinion: no agent wallet is declared, or gross income is under 50 USDG, or there are fewer than 10 incoming transfers, or data the method requires is unavailable. | Default thresholds, version 0.1. It is decided before the other three. |
Pledge
- Method
| Term | Definition | Test |
|---|---|---|
Kept | The check found that what the pledge states, as its template defines it, held for the period checked. | Checks run weekly, and each result carries the evidence that decided it. Code decides the result from the test of the template. No model does. |
Broken | The check found evidence that what the pledge states did not hold. | The result carries the evidence that decided it. Each of the seven templates of version 1 defines its own test. |
Pending | The check cannot be decided yet: the date or the period the pledge names has not been reached. | The pledge is checked again each week. |
Unverifiable | The check cannot be decided from public data: what it needs cannot be read from the chain or from a public source. | The result names the data that was missing. |
Status
- Method
| Term | Definition | Test |
|---|---|---|
Chartered | The agent has filed, and every rule of the rulebook version in force passes. | Each rule is a check that code runs, and the status can be reproduced from the rulebook version and the evidence alone. At least 30 days of history is needed first. |
Provisional | The filing of the agent was accepted, and the rules are being observed. | At least 30 days of history is needed before every rule can pass. |
Unchartered | Unchartered means the agent has not filed. The term carries no judgement. | It is the default for every agent. When an operator withdraws a filing, the agent returns to it and the history stays. |
Revoked | The agent was Chartered, and then a rule failed and stayed failed past the grace period, or a statement in its filing was shown to be false. | Default grace period: 14 days. A revocation names the rule and links the evidence. |
Breach
- Method
| Term | Definition | Test |
|---|---|---|
Tested | The operator opened a public round against the agent's own declared endpoints and wallets, and the round stayed open for at least 30 days with the prize locked throughout. | Default threshold from the specification: 30 days. |
Not tested | No round that meets the test for Tested is on record for the agent. The term carries no judgement. | It is the default, and it holds for an agent that never ran a round. The answer gives the reason in one sentence. |