Skillware skills · security/prompt_injection_firewall
Prompt Injection Firewall
Offline deterministic pre-flight scanner for hostile instructions in untrusted text — hidden HTML, Unicode smuggling, nested encodings, and instruction overrides — before content reaches an LLM. No auditing model, no API keys.
Capabilities
- Offline scan
- Sanitized output
- Structured findings
Problems this skill solves
People search for tasks — not always "skillware":
- prompt injection firewall agent skill
- scan untrusted text before llm
- jailbreak detection python agent
- sanitize scraped content for ai agent
- hidden html prompt injection detector
- offline prompt injection scanner
Install
pip install skillware then SkillLoader.load_skill("security/prompt_injection_firewall")
FAQ
- How do I scan untrusted text for prompt injection before an LLM call?
- Use Skillware security/prompt_injection_firewall — pip install "skillware[security_prompt_injection_firewall]", load the skill, pass source_text; it returns is_safe, findings, and optional sanitized_text locally.
- Does the firewall call an LLM to audit content?
- No. It is pattern-based and deterministic — no network, no API keys, no auditing model in the loop.