Pre-Flight Prompt Scrubber
Browser-and-CLI plug-in that scans the prompt and context you are about to send to an AI assistant and blocks or replaces anything that looks like a secret, customer record, or proprietary file path.
Individual developers and data scientists who paste real code and data into AI chat tools every day
- Pattern-based detection of API keys, tokens, customer emails, and internal hostnames across pasted snippets
- Per-vendor policy: stricter redaction for consumer logins than for enterprise API keys
- One-click 'safe paste' mode that replaces secrets with placeholder tokens before sending
- Audit log of what was redacted so you can review and tune the rules over time
Public wire-level proof that mainstream AI CLIs upload entire repos and secrets by default has made developers wary of pasting real codebases into any agent without a local guardrail.
Wire-level Grok CLI analysis is widely covered and discussed (Hacker News thread referenced in MaydayV/grok-upload-audit), and there are active developer conversations about prompt sanitization, indicating real, recent concern.grok-upload-audit/README.md (references Hacker News discussion of wire-level analysis) ↗What xAI Grok Build CLI actually sends to xAI — a wire-level analysis ↗
The space is already crowded with near-identical offerings: LLM Secret Guard (Chrome extension covering ChatGPT/Claude/Gemini/Grok), EraseAI browser firewall, prompt-sanitizer on PyPI, LeakGuard MVP, LLM-Privacy-Guard, and Prompt Firewall variants — direct differentiation will be hard.LLM Secret Guard - ChatGPT & AI Secret Leak Protection ↗EraseAI — AI Firewall for ChatGPT, Claude, Gemini & LLM Apps ↗ai-prompt-sanitizer · PyPI ↗
Security tooling can command prices (enterprise DLP benchmarks at $15-$30/seat/month) and the Prompt Firewall site explicitly targets individuals/teams as a cheaper alternative, but the target user is price-sensitive developers and many free extensions/OSS options exist, capping realistic ARPU.Prompt Firewall — Use AI without leaking your life into it ↗
Regulatory pressure (EU AI Act, GDPR, sectoral data laws) plus recurring CLI/agent data-exposure incidents make pre-flight scrubbing a structural, not faddish, need; trend has multi-year legs.Grok Build Repository Upload Allegations Explained ↗
MVP is a browser extension plus CLI shim using well-known libraries (detect-secrets, trufflehog-style regex/entropy) and existing open-source prompt-sanitizer pipelines — buildable by a single developer in weeks.GitHub - jeslor/prompt-sanitizer ↗ai-prompt-sanitizer · PyPI ↗