What problem does it solve? Products claiming zero-knowledge or end-to-end encryption may still transmit plaintext content, keys, or recovery secrets over the network. This Skill verifies those claims by planting synthetic sentinel values in test flows and scanning captured HAR traffic to detect where protected data actually appears. ## Core Features & Use Cases - Sentinel-Based Leak Detection: Generate high-entropy synthetic sentinels across categories (plaintext content, metadata, encryption keys, recovery secrets, credentials) and detect them in URLs, headers, cookies, request/response bodies, and telemetry. - Recursive Decoding: Automatically decode URL encoding, base64, base64url, JWT segments, hex, gzip/zlib, and nested JSON to catch encoded plaintext that naive scans miss. - Structured Findings: Emit schema-shaped findings with ZK-NET IDs, severity, classification labels (third-party payload, telemetry, unencrypted transport), redacted evidence excerpts, and retest steps. - Use Case: Auditing a notes app that claims end-to-end encryption: generate sentinels, exercise login/sync/share flows while capturing HAR, then run the scanner to confirm whether note bodies or keys ever reach the server or third-party analytics in readable form. ## Quick Start Ask the agent to audit the supplied HAR capture with its sentinel manifest using the network zero-knowledge test and report any leaked plaintext findings with severity and evidence.
