Credit Scorecards

Three clean receipts still score as probationary

The captured agent holds three allowed receipts on one tool server, ledger, each for settle_invoice with a governed ceiling of 2,500 minor units ($25), plus one underwriting decision. The export reads them from the receipt store and signs the result with the authority seed.

$ chio --receipt-db ./receipts.db trust credit-scorecard export \
    --agent fixer \
    --authority-seed-file ./authority.seed

schema:                 chio.credit.scorecard.v1
generated_at:           1788542733
signer_key:             fb4d4902815fd5db1e628702eeb87c48271872ac2fde57298767c3f8c04f9402
subject_key:            3b9ed9e1a200eec3b3fcc461f54ca1cf8af94e7ab205751d0d330320afedcb41
overall_score:          0.6000
confidence:             Low
band:                   Probationary
probationary:           true
matching_receipts:      3
matching_decisions:     1
anomaly_count:          1

You should see eleven aligned lines: the schema, the generation time, the key that signed the export, the subject, and the summary fields that decide the band. Three receipts clear neither target_receipt_count nor target_span_days, so probationary is true, confidence is Low, and the band is Probationary even though every receipt settled cleanly. The single anomaly is the one low confidence raises.

$ chio --receipt-db ./receipts.db receipt list --admin-all \
    | jq -r '[.decision.verdict, .tool_server, .tool_name, .metadata.governed_transaction.max_amount.units] | @tsv'

allow	ledger	settle_invoice	2500
allow	ledger	settle_invoice	2500
allow	ledger	settle_invoice	2500

$ chio --receipt-db ./receipts.db --json trust credit-scorecard export \
    --agent fixer \
    --authority-seed-file ./authority.seed \
    | jq '.body.summary'

{
  "matchingReceipts": 3,
  "returnedReceipts": 3,
  "matchingDecisions": 1,
  "returnedDecisions": 1,
  "currencies": [
    "USD"
  ],
  "mixedCurrencyBook": false,
  "confidence": "low",
  "band": "probationary",
  "overallScore": 0.6,
  "anomalyCount": 1,
  "probationary": true
}

Confidence caps what a strong record can earn

Every scorecard resolves to one of five ordinal bands, from most permissive to most restrictive: Prime, Standard, Guarded, Probationary, Restricted. A band is a categorical output. It follows from the reason codes, the anomalies, and the confidence the evidence supports, and chio-credit encodes no score threshold per band.

The reasoning is spread across four scored dimensions. Each carries an optional score from 0.0 to 1.0, a weight, a description, and evidence references back to the receipts, decisions, or attestations behind it:

• ReputationSupport scores the subject's composite reputation, attenuated by imported-trust dependencies.
• SettlementDiscipline weighs pending against failed settlement counts on the subject's receipts.
• LossPressure measures provisional loss amounts still outstanding against the subject.
• ExposureStewardship reads concentration, currency mix, and reserve-aware utilization.

Sample size sets confidence (Low, Medium, or High), and confidence caps the band. An agent with a strong reputation score and 12 receipts cannot reach the band of an agent with the same score and 12,000. The subject stays probationary while receipt_count < target_receipt_count or span_days < target_span_days, and the reason codes SparseReceiptHistory and SparseDayHistory name the target it misses.

Read more in Reputation Scoring, which builds the composite behind ReputationSupport.

A critical anomaly outweighs a clean reputation

An anomaly is a typed reason the band would otherwise sit too high for the underlying signal. It reuses the reason-code enum and adds a severity of Info, Warning, or Critical, with a description and its own evidence references.

Severity moves the band. A scorecard with a strong composite score and a FailedSettlementBacklog anomaly at Critical shifts toward Restricted, however clean the reputation looks. The other reason codes cover the remaining pressures: pending settlement backlog, provisional loss, a mixed-currency book, low reputation, imported trust, and missing decision coverage.

The export refuses to sign a scorecard it cannot ground

A scorecard needs a signing key, receipts for the subject, and money on those receipts. A missing prerequisite exits 1 with an empty stdout instead of producing an empty scorecard. Without a signing key:

$ chio --receipt-db ./receipts.db trust credit-scorecard export --agent fixer

error[CHIO-CLI-SIGNING-KEY-REQUIRED]: the export is signed, and no signing key was given
  help: pass --authority-seed-file <path> or --authority-db <path>
  code: urn:chio:error:cli:signing-key-required
  docs: https://chio.computer/docs/reference/errors#error-cli-signing-key-required

With a key, against a store that holds no receipt for the subject:

$ chio --receipt-db ./empty.db trust credit-scorecard export \
    --agent fixer \
    --authority-seed-file ./authority.seed

error[CHIO-ECONOMY-SCORECARD-EMPTY]: no governed receipt in the store matches the selected agent
  help: point --receipt-db at the store that holds the agent's governed receipts, or run governed calls as the agent first
  code: urn:chio:error:economy:scorecard-empty
  docs: https://chio.computer/docs/reference/errors#error-economy-scorecard-empty

The third gate needs monetary exposure. credit_scorecard_position_denominator returns a value only when governed exposure, or settled plus pending plus failed units, is above zero. A subject whose receipts carry no governed_transaction.max_amount and no charged cost has receipts to count and nothing to score, and the export answers credit scorecard requires monetary exposure in the requested window.

The local reputation scorer returns Unknown metrics against an empty store. The credit export signs nothing at all, so every scorecard an operator reads rests on governed spend.

The band sets a facility's limit and terms

A scorecard is an evaluation report and mints no capital. The credit facility report carries its summary in a scorecard field, derives a disposition from it, and builds terms only when that disposition is Grant:

• Deny when the band is Restricted, the subject misses the minimum runtime assurance tier, or the tool server requires certification and none is active.
• ManualReview when the book is mixed-currency, the scorecard carries FailedSettlementBacklog or PendingSettlementBacklog, the subject is probationary, or runtime-assurance evidence came from more than one verifier family.
• Grant otherwise.

The credit limit is the subject's own exposure scaled by the band and then by confidence, and floored: floor(base_units × band_factor × confidence_factor). The confidence factor is 1.0 for High, 0.9 for Medium, and 0.75 for Low.

For a Probationary band the terms builder returns None, and the disposition is ManualReview regardless. A Restricted band is denied. The two weakest bands also demand the strongest attestation: credit_facility_minimum_runtime_assurance_tier asks RuntimeAssuranceTier::Attested of Prime, Standard, and Guarded, and Verified of Probationary and Restricted. A thin history without verified runtime attestation is denied on that prerequisite before the band is weighed.

Terms carry their own ttl_seconds, which expires a granted facility. A report without terms falls back to seven days for Grant and ManualReview and one day for Deny. Every granted facility is stamped CreditFacilityCapitalSource::OperatorInternal: the capital behind it is the operator's own.

Read more in Credit Facilities, which follows a granted facility through allocation, bonds, and loss.

Next: turn a band into a decision