Skip to content
ARCHYTASAI

TASK-FIN-104-04 · Task family

Mandate breach investigation & resolution. Breach diagnosis, corrective proposals and escalation.

Custom buildenvironment · ENV-FIN-104 Portfolio Risk & Mandaterole · Investment Compliance Officer

Overview. Classify each breach, propose remedies consistent with policy and escalate with the evidence.

Post-trade compliance flags two breaches. One came from a market move, the other from a trade. Passive and active breaches are handled differently, and the agent has to classify each one, propose remedies and escalate.

Selling immediately at any price isn't always required for a passive breach, and an active breach can't be explained away as a market move.

Value to the buyer

Breach handling depends on the cause. Trains agents to diagnose before they remedy.

Specification. The task family, field by field.

spec · task/task-fin-104-04.yamlENV-FIN-104
task_id:
TASK-FIN-104-04
domain:
Portfolio Management & Investment Compliance
parent_environment:
ENV-FIN-104 · Portfolio Risk & Mandate
target_role:
Investment Compliance Officer
objective:
Classify each breach, propose remedies consistent with policy and escalate with the evidence.
starting_context:
The post-trade compliance report, trade blotter, price history, mandate and breach policy.
tools:
  • get_positions
  • get_market_prices
  • get_investment_mandate
  • simulate_trade
  • check_pretrade_compliance
  • request_exception_approval
required_actions:
  • Determine each breach's cause from trades and prices
  • Classify each as active or passive
  • Propose remedies within the policy's time limits
  • Escalate to compliance and the PM
expected_outputs:
  • Breach analysis
  • Remediation proposal
  • Escalation record
interaction:
Agent plus simulated compliance and PM approvers.
verification:
Classification, remedies and timelines are compared with the breach policy and the seeded cause.
reward:
Misclassifying an active breach as passive scores zero. Credit for each correct classification and remedy.
difficulty:
Calibrated against reference models during acceptance testing
task_horizon:
Multi-step investigation with escalation. Execution limits are agreed with the buyer.
delivery:
Task instances with seeded breaches, reference analyses and verifier code.
commercial_unit:
Per validated task instance

Sample prompt. What the agent is asked to do, as a buyer would see it.

prompt · task-fin-104-04Example
Post-trade compliance flags two breaches. Issuer Velan is at 10.4% against a 10% limit after its share price rose 6% this week, with no Velan purchases. A trader bought shares of a company on the restricted list yesterday. The breach policy requires passive breaches to be remedied within ten business days, in the fund's interest, and active breaches to be reversed promptly and reported. Investigate, propose remedies and escalate.

Verification criteria. What the verifier checks, and how.

  • Velan classified

    Passive breach: the weight rose from price moves, not trades.

  • Velan remedy

    Reduce below 10% within ten business days, scheduled to limit market impact.

  • Restricted trade classified

    Active breach caused by a trade.

  • Restricted remedy

    Reverse the trade promptly and report it under the policy.

  • Escalation

    Both breaches go to compliance and the PM with the evidence.

Key failure modes. Where agents go wrong on this task.

  • Selling Velan immediately regardless of market impact
  • Treating the restricted-list purchase as a passive breach
  • Reporting the breaches without remedies

Other task families. The rest of Portfolio Risk & Mandate, on the same FinanceWorld state.

Every environment is built to order. Tell us which workflow you need to train or evaluate, and where you are in the process.

Request sample tasksDiscuss custom developmentRequest technical integration detailsSee ways to buy