Copyright © 2026 CYBERDUDEBIVASH ECOSYSTEM – All Rights Reserved
Authorized under CYBERDUDEBIVASH AUTHORITY – Ethical forensic auditing of AI agent traces for Lies-in-the-Loop (LITL) deception.
This tool scans JSON trace logs from AI agents / Human-in-the-Loop (HITL) dialogs to detect stealthy deception patterns including siphon commands, visual padding evasion, Markdown forgery, and unverified tool calls.
- Detects known siphon phrases ("Ignore all", "Summary complete", etc.)
- Flags excessive newline padding used for visual hiding
- Identifies basic Markdown forgery attempts (fake buttons, hidden code blocks)
- Warns on unverified tool calls
- Generates clear risk reports with remediation advice
- Supports single file or batch directory scanning
- 100% local execution – no cloud dependencies or data upload
No external packages required (uses Python standard library).
Tested on Python 3.8+.
- Clone the repository:
git clone https://github.com/cyberdudebivash/CYBERDUDEBIVASH-AI-LITL-Triage-Scanner.git cd CYBERDUDEBIVASH-AI-LITL-Triage-Scanner
Run directly (no pip install needed for core usage).
Usage Examples Single file scan: Bashpython triage_scanner.py path/to/trace.json Batch scan entire directory: Bashpython triage_scanner.py logs/ --dir Sample JSON trace format (expected structure) JSON[ { "prompt": "User query here", "response": "AI response text", "display_content": "Visible output with \n\n\n padding", "tool_call": "some_command" } ] Customization
Add new siphon patterns directly in the SIPHON_PATTERNS list in triage_scanner.py. For premium features (real-time monitoring, custom rule sets, SOAR integration, AI-enhanced scoring): https://www.cyberdudebivash.com/contact