Risk scoring

Skill yogeshg665/sleuth-fraud-investigator/skills/risk-scoring

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 risk-scoring

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

Aggregates all triggered risk signals into a single explainable score from 0 to 100 using confidence weighting, a corroboration boost, and a critical-signal floor. WHEN: "score the risk", "calculate fraud score", "aggregate signals", "compute risk score", "combine fraud indicators".

SKILL.md

2.0 KB, 393 tokens by cl100k_base, as published. Nobody here has run it

Risk Scoring

Overview

Combines the individual risk signals into one aggregate score. The method blends signal severities by their configured weight and confidence, applies a small boost when multiple independent signals corroborate, and never lets the score fall below a definitive critical signal.

When to Use

After all detection skills have run and before fraud-decisioning.

Inputs

Input	Required	Description
`case.signals`	yes	The signals emitted during detection.
`skill_weights`	yes	Per-skill weights from configuration.

Process

Select the triggered signals. If none, the score is zero.
For each signal, compute a weight equal to its configured skill weight times its confidence.
Compute the weighted average of severities.
Apply a corroboration boost that grows with the number of independent triggers, capped to avoid runaway inflation.
Compute the critical floor as the maximum severity among critical signals.
The final score is the greater of the boosted blend and the critical floor, bounded to the range 0 to 100.

Outputs

A risk_score on the case.

Reference Implementation

src/fraud_investigator/agents/risk_scoring_agent.py.

Rationalizations

Excuse	Rebuttal
"Just take the maximum severity."	A pure maximum is brittle to a single noisy signal; blending is more robust.
"Ignore the critical floor to smooth results."	Definitive indicators must not be diluted; the floor is intentional.

Red Flags

The score can be lower than a confirmed deny-list signal.
Weights are applied without confidence.

Verification

A case with a critical deny-list signal scores at or above that signal's severity.