Production pilot · authorized data only

Personal Exposure Defense

Discover exposure. Verify evidence. Score risk. Create and track remediation work. Get human approval before external action. Prove progress.

Start or return Management portal

Operating system

One case loop, visible proof, guarded action

Cloak Harbor turns authorized identifiers into evidence, risk, work, approvals, reports, and recurring review. No external action is taken unless a protected server route and a human approval allow it.

Capabilities

Human gates

External action

false

Capability loop

Catalog coverage by operating stage.

No external action

Intake

Confirm who is protected, what is in scope, and what the system is allowed to touch.

Discover

Plan bounded exposure discovery against authorized identifiers.

Verify

Review evidence before anything becomes a risk or action item.

Score

Translate exposure into a clear StealthScore and explain what would improve it.

Queue

Turn validated findings into accountable work with owners and due dates.

Operate

Run the work through Kanban, service setup, partner handoffs, and communications drafts.

Cadence

Keep the 30-day remediation rhythm visible for clients and operators.

Draft

Draft opt-out, correction, partner, and client packets without sending them.

Approve

Require a person to approve sensitive work before any outside-world touch.

Prove

Package the before/after story into reports, screenshots, and audit evidence.

Monitor

Watch for re-exposure after cleanup without pretending live monitoring is active.

Support modes

Postgres and Vercel support are explicit. Missing infrastructure degrades honestly.

Postgres

Vercel

Redacted public baseline

Connected

Public-safe baseline only; saved client records replace it as soon as a real authorized case is selected.

PED OS capability catalog

Connected

Catalog remains readable even when persistence is unavailable.

Postgres capability support flags

Configured

Vercel reports redacted/degraded mode only until DATABASE_URL is configured.

Local audit event persistence

Configured

Deployments without DATABASE_URL show public-safe trace summaries and do not claim durable audit persistence.

Protected capability bootstrap API

Server-authorized

Without Postgres, the route returns postgres_not_configured instead of fake success.

MCP capability introspection

Connected

MCP reports public-safe degraded runtime when Postgres is absent.

Sensitive action matrix

Actions stay explainable, gated, and auditable.

Client onboarding

Human review

Authorized or redacted data only

externalActionTaken:false

Customer onboarding

Waiting client

No document upload success is claimed

externalActionTaken:false

Management onboarding

Planned

No team changes are saved

externalActionTaken:false

Discovery planning

Planned

No live scan has been performed

externalActionTaken:false

Calendar cadence

Server-authorized

No real calendar event is created

externalActionTaken:false

Human review center

Human review

No auto-submit

externalActionTaken:false

Action packet studio

Dry-run

AI drafted, human approval required

externalActionTaken:false

Communications hub

Planned

No sending without protected server action

externalActionTaken:false

Executive landing

Preview

A boardroom-ready first screen that explains Personal Exposure Defense as one practical operating loop, not a generic dashboard.

Open surface

Client onboarding

Human review

A calm intake wizard that explains what happens next and makes consent boundaries clear before any discovery work.

Open surface

Discovery planning

Planned

A planning console for bounded discovery that shows what would be searched before any live scan exists.

Open surface

Evidence review

Connected

An evidence vault that makes clear what was captured, why it matters, and what still needs human review.

Open surface

Defense loop ledger

Evidence-to-action proof chain

Every stage declares what is real, what is gated, and whether anything touched the outside world. Checkpoint persistence can be recorded through POST /api/v1/loop when Postgres and server authorization are configured on Vercel or local Docker.

Stages

Findings

Actions

externalActionTaken:falseCheckpoint persistence: PostgresVercel-ready protected routePOST /api/v1/loop

Principal

Redacted

Each case starts with an authorized client, explicit identifiers, consent, and scope before any discovery work begins.

Open stage

Discovery

Planned

The discovery stage shows what would be searched and why, while refusing to run a live people-search crawl.

Open stage

Evidence

Connected

Each exposure has evidence with attribution, confidence, redaction state, and trace linkage before it can drive action.

Open stage

Risk

Connected

StealthScore turns the evidence set into a plain before/after risk posture with transparent drivers.

Open stage

Work

Connected

The risk posture becomes owned work, with every task linked back to an exposure and lane.

Open stage

Cadence

Connected

The calendar makes the 30-day defense rhythm visible: immediate hygiene, partner review, re-checks, and reporting.

Open stage

Action

Dry-run

Action packets are drafts only: they help a human decide what to send, but they do not touch outside systems.

Open stage

Approval

Human review

The review center separates useful AI/operator drafting from the authority to act outside Cloak Harbor.

Open stage

Report

Ready for review

The report compresses the case into an executive-ready proof package: status, risk, actions, open issues, and limits.

Open stage

#10

Recurrence

Degraded

The loop does not end at a report; recurring exposures are watched and routed back into evidence, work, and review.

Open stage