Module 1 · Failure Radar

About ScreenPath

An open-source experiment exploring whether external operational failure signals can become useful inputs into financial-services control assurance.

The thesis

Learn from failures outside your walls before reproducing them inside your own.

Module 1 converts governed external discovery into evidence-grounded FailureShapes and themes. Future modules connect those patterns to enforcement, policy change, internal controls and deterministic non-production assurance.

This is a public preview. It serves a captured, read-only snapshot of investigations that really ran against live sources. Starting new investigations and scheduled monitoring are disabled here so that anonymous visitors cannot spend search, regulator-API or model budget — both run in a self-hosted deployment.

Truth model

ScreenPath keeps these four states apart on purpose. Collapsing them is how an intelligence product starts claiming things it has not established.

DISCOVERED
A source says this exists. ScreenPath holds the metadata — title, URL, domain, rank — and nothing more.
OBSERVED
ScreenPath was permitted to acquire the content, and holds it with its provenance.
INFERRED
An interpretation drawn from observed evidence. Always attributed, never presented as proof.
PROVEN
A deterministic assertion engine executed a test and recorded the outcome. Module 1 never claims this.

A model interprets evidence. A model never determines PASS, FAIL or compliance. Every model assessment must validate against a canonical schema, agree with its own extracted features, and quote the evidence verbatim — or it is discarded. Deterministic proof belongs to a later module and is not claimed anywhere in Module 1.

Roadmap

Module 1Failure RadarWhat externally reported operational contradictions are visible?LIVE
Module 3EnforcementWhere have regulators acted on similar failure mechanisms?PLANNED
Module 4Policy WatcherWhat changed in external rules and policies?PLANNED
Module 5Control Map + Change ImpactWhich internal systems, processes and controls could be affected?PLANNED
Module 6Assurance LabCan the analogous failure be reproduced in non-production?PLANNED
Module 7Failure MemoryWhat did we previously learn and fix, and could it regress?PLANNED
Module 8CommandOne surface connecting intelligence, change and assurance.PLANNED

Known limitations

  • Public search providers frequently do not expose a trustworthy publication date. Undated evidence is deliberately excluded from current-period incident claims rather than assumed recent.
  • Investigations are anchored to an approved institution registry. Source applicability, alias matching and CFPB coverage are reference data, not runtime guesses.
  • A local model host is supported for private self-hosted use, but is never part of the hosted runtime.
  • Modules 2–8 are not built. Nothing in this product proves a control passes or fails.

Built in public

Fork it, test the assumptions, open an issue, propose a failure family, improve a provider, or contribute a module.

View on GitHub ↗