Tools / Governance

detect_leakage()

Key

Checks every item pinned to a decision: anything with available_at later than the decision's as_of is a leak — knowledge from the future dressed up as hindsight — and comes back as a violation report (invariant I2).

detect_leakage(decision_id)→ VIOLATION REPORT

Callable over MCP as detect_leakage(decision_id). REST mirror POST /v1/leakage Planned — alpha is MCP-only

Why your agent calls it

The hindsight detector. Any pinned item whose available_at postdates the decision's as_of is knowledge from the future — a backtest lying to you. One call returns every violation, or the clean bill that makes the decision's record trustworthy.

What it returns

detect_leakage() · Catch any future fact that leaked into the past.ILLUSTRATIVE

Call it from an agent

Over MCP, your agent invokes it on its own when the moment calls for it — just describe the situation. A Python SDK for headless runtimes is planned; this is what the call will look like.

agent · python sdkSDK · PLANNED — ALPHA IS MCP-ONLY

from cascadian import Cascadian

cx = Cascadian(api_key=os.environ["CASCADIAN_API_KEY"])

result = cx.detect_leakage(
    decision_id="dec_0142",
)

Parameters

Param	Type	Description
decision_idREQUIRED	string	The decision to audit for leakage.

Example response

detect_leakage()JSON · ILLUSTRATIVE

{
  "decision_id": "dec_0142",
  "violations": [],
  "checked": "every pinned item: available_at ≤ as_of"
}

Related — Governance

scopedrequest_approval()Ask the human. The token never crosses this boundary.keyget_pending_approvals()What is waiting on a human.keyreconstruct_decision()Re-run the pinned reads. The diff is the audit.publiccapabilities()The honesty endpoint — what is real vs planned.