Skip to the rows

agentsclimarketplace

Fraud decisioning

Skill yogeshg665/sleuth-fraud-investigator/skills/fraud-decisioning

Sleuth - Agent Skills that automate payment fraud investigations, plus a deterministic Python engine as the executable reference. Explainable, deterministic scoring and decisions with tokenized card references and human-review gates.

Install

npx -y skills add yogeshg665/sleuth-fraud-investigator --skill fraud-decisioning

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Maps the aggregate risk score and signals to an explainable recommendation of approve, escalate, or decline, with a confidence value and recommended actions. WHEN: "make a fraud decision", "approve or decline", "should I block this", "decision policy", "escalate to review", "final fraud verdict".

SKILL.md

2.1 KB, 399 tokens by cl100k_base, as published. Nobody here has run it

Fraud Decisioning

Overview

Applies the configurable decision policy to the scored case and produces a recommendation with a confidence value, the reasons behind it, and the actions to take next.

When to Use

After risk-scoring, once the case has an aggregate score.

Inputs

Input	Required	Description
`case.risk_score`	yes	The aggregate score.
`decision_policy.decline_threshold`	yes	Score at or above which to decline.
`decision_policy.escalate_threshold`	yes	Score at or above which to escalate.

Process

If the score is at or above decline_threshold, recommend decline and list blocking and notification actions.
Else if the score is at or above escalate_threshold, recommend escalate to manual review and request step-up authentication.
Otherwise recommend approve with continued passive monitoring.
Derive confidence from the margin between the score and the relevant threshold and from signal agreement.
Attach the triggered signal rationales as the reasons for the decision.

Outputs

A Decision with outcome, confidence, reasons, and recommended actions.

Reference Implementation

src/fraud_investigator/agents/decision_agent.py.

Rationalizations

Excuse	Rebuttal
"Auto-decline borderline cases to be safe."	Borderline cases belong in the escalate band for human review.
"Skip the reasons to save space."	A decision without cited reasons is not auditable and must not ship.

Red Flags

A decline is issued with no triggered signals.
Thresholds are inverted so that low scores decline.

Verification

The outcome is consistent with the score and the configured thresholds, and the decision lists its reasons.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.