GATEKEEPPROMPT FIREWALL · LIVE DEMO

GITHUB ↗

Gatekeep is a drop-in reverse proxy for the Anthropic API. It scans every prompt before the model sees it: secrets are blocked, personal data is redacted, everything is audit-logged without ever storing the raw text. This page runs the same detection and policy pipeline, ported to JavaScript, entirely in your browser. No API key, no server, nothing leaves this page.

On load, this page replayed the repo's red-team corpus (harness/corpus.jsonl, the same file harness/run_harness.py scores) through the in-browser engine: 30 dirty prompts and 20 clean traps built to bait false positives (git SHAs, UUIDs, Luhn-invalid cards, invalid SSNs, code). Scoring is strict: the decided action must equal the expected action exactly. The JS port is also parity-tested against the Python engine finding-by-finding in CI style via demo/parity.test.js.

01 REQUEST
client requests model:
SAMPLES, STRAIGHT FROM THE RED-TEAM CORPUS
Ctrl+Enter
02 CHECKPOINT DECISION
NO REQUEST INSPECTED YET · send a prompt or hit a sample
03 WHAT THE MODEL SAW
awaiting a request…

04 AUDIT TRAIL
No decisions yet this session.

One row per decision, the same columns as the proxy's SQLite audit table. The raw prompt is never stored: only a SHA-256 of it, the finding categories and detector names, the routing outcome, and the decision latency. That is the never-log rule, and it holds even here, in a page that never leaves your machine.