Nova-HuntingNova-HuntingOfficialยท1 Agent Skills Included

nova-tracer

Prompt injection detection and session monitoring for coding sessions

Detects prompt injection attacks in files, web pages, and command output using keyword, semantic, and LLM scanning tiers. Blocks dangerous commands like recursive deletes before execution and warns about suspicious content after it is read. Generates interactive HTML reports with timelines, verdicts, and summaries so you can audit every session without manual log review.
npx skills add Nova-Hunting/nova-tracer --all -g -y

All Skills in This Repository (1)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install nova-tracer?โ–ผ

Run `npx skills add Nova-Hunting/nova-tracer --all -g -y` in your terminal to install it globally. Then restart Claude Code so the monitoring hooks activate.

How to detect prompt injection in Claude Code?โ–ผ

Nova-tracer scans tool output with three tiers: fast keyword matching, ML-based semantic similarity, and optional LLM evaluation. When a threat is found, it warns Claude to treat the content with suspicion.

Can nova-tracer block dangerous commands?โ–ผ

Yes. Its pre-tool hook actively blocks destructive operations like rm -rf, fork bombs, and credential exfiltration before they execute.

Does nova-tracer work without an API key?โ–ผ

Yes. Keyword and semantic detection run locally with no API key. The LLM tier and AI summaries require an Anthropic or OpenAI key, but can be disabled in the config.

Can I add custom detection rules?โ–ผ

Yes. Create .nov rule files in the rules directory using the YARA-inspired syntax with keywords, semantic descriptions, and LLM questions.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’