AgentSkillsCN

credit-risk

信用风险工作流，适用于违约概率、违约损失以及违约时暴露额的估算，并结合投资组合层面的损失聚合。适用于在涉及预期损失建模、尾部损失压力测试、债务人层级风险评分，或生产信用额度控制的任务。

SKILL.md
--- frontmatter
name: credit-risk
description: "Credit risk workflows for probability of default, loss given default, and exposure at default estimation with portfolio-level loss aggregation. use when tasks involve expected-loss modeling, tail-loss stress testing, obligor-level risk scoring, or production credit limit controls."

Credit Risk

objective

Measure and control credit loss risk across obligors, sectors, and portfolios with calibrated pd/lgd/ead models.

workflow

  1. •define obligor universe, default definitions, and risk horizon conventions.
  2. •estimate pd, lgd, and ead with segment-aware calibration.
  3. •aggregate expected and stressed losses to portfolio and desk levels.
  4. •validate calibration, discrimination, and tail-loss behavior.
  5. •deploy only when risk metrics and controls are stable and explainable.

required validation

  • •expected-loss decomposition consistency across pd, lgd, and ead.
  • •calibration drift by rating band and sector.
  • •default-event capture and false-alarm behavior.
  • •stressed loss sensitivity under macro and spread shocks.
  • •concentration risk by counterparty and sector cluster.

risk controls

  • •enforce obligor and sector concentration limits.
  • •enforce threshold alerts for pd and expected-loss jumps.
  • •enforce model fallback when calibration quality degrades.

outputs

  • •run python scripts/credit_risk_validation.py input.csv --output validation.json and keep the json artifact.
  • •write an implementation memo using references/credit-risk-playbook.md with assumptions, tests, limits, and rollout plan.

resources

  • •use scripts/credit_risk_validation.py for deterministic validation.
  • •use references/credit-risk-playbook.md for the domain checklist and delivery structure.