Prompt Injection Dataset Scanner v2

Scan public pages for prompt-injection phrases before they are indexed into RAG or used by agents.

Scan AI agent prompts, skill files and manifests for prompt injection, secret patterns, risky permissions and dependency signals. Get an explainable JSON report with redacted evidence. Rule-based checks; no input execution or external AI model.

Scan webpages or text for prompt injection, secret exfiltration, hidden instructions and Unicode control characters.

Prompt Injection Dataset Scanner

zentrafoundry/zentra-prompt-injection-quarantine

Scan and sanitize dataset records before they enter LLM, RAG, or agent pipelines.

Stop prompt-injection and jailbreaks before your agent reads them. Scans tool results, web pages, and user input for instruction-override, hidden Unicode tag-smuggling, ChatML token injection, and 30+ attack patterns. Returns a risk score, flags, and sanitized text.

Heuristically scan untrusted text and public web pages for prompt-injection patterns, flag suspicious instructions, and separate them from retained content for safer AI-agent workflows.

Screen text, HTML, and URLs for prompt injection and unsafe agent instructions.

Prompt Injection & Jailbreak Payload Corpus

jungle_synthesizer/prompt-injection-jailbreak-payload-corpus-scraper

Normalizes published prompt-injection and jailbreak datasets from HuggingFace and GitHub research repos into one labeled corpus: technique, target model, defense bypassed, license, cross-source dedup. Defensive only — aggregates public data for guardrail/eval testing, never targets a live LLM.

Agent Skill Security Scanner

scraper_guru/agent-skill-security-scanner

Scan Claude, Codex, Cursor, and skills.sh agent skills for prompt injection, secret access, unsafe commands, exfiltration, and supply-chain risks.

Convert webpages to compact Markdown with provenance, token estimates and prompt-injection risk flags for AI agents.