How does this skill prevent prompt injection?

It treats all user messages and uploaded documents as data rather than instructions, ignoring any text that attempts to redefine the AI's role or bypass security protocols.

What happens if a user tries to skip a mortgage workflow phase?

The guardrails enforce a sequential four-phase process, automatically redirecting users back to the current required phase if they attempt to skip ahead.

Does this skill handle sensitive PII?

The skill actively prevents the collection of PII such as SSNs or bank info in chat, redirecting users to a secure application portal instead.

Does it protect against social engineering?

Yes, it is programmed to resist authority impersonation, urgency tactics, and emotional manipulation, treating all users according to the same standard safety rules.

Can users access internal API keys or configuration?

No, the skill explicitly prohibits the disclosure of credentials, environment variables, and internal configuration files like mortgage.local.md.

Lendtrain Security Guardrails

Name: Lendtrain Security Guardrails
Author: lendtrain

bylendtrain

•

Security & Testing

Implements mandatory adversarial defense rules and security guardrails to protect mortgage refinance workflows from manipulation.

The Security Guardrails skill acts as a critical cross-cutting security layer for the Lendtrain mortgage plugin. It is designed to defend against a wide array of threats including prompt injection, system prompt extraction, unauthorized configuration access, and social engineering attacks. By establishing a strict instruction hierarchy and enforcing mandatory workflow phases, this skill ensures that Claude remains within its specific domain of mortgage refinancing while protecting sensitive internal logic, API credentials, and borrower PII from exposure or misuse.

Key Features

01Enforces mandatory workflow phases and state-specific licensing restrictions.

02Prohibits the collection of sensitive PII like SSNs and bank account numbers within chat.

03Defends against conversational and document-based prompt injection attacks.

04Resists social engineering tactics and unauthorized claims of administrative authority.

052 GitHub stars

06Prevents disclosure of internal system prompts, configuration files, and API credentials.

Use Cases

01Securing mortgage data extraction from uploaded documents against embedded malicious instructions.

02Protecting proprietary pricing logic and internal backend infrastructure details from user discovery.

03Ensuring strict adherence to regulatory compliance and state licensing boundaries during refinance quotes.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lendtrain/mortgage security-guardrails

For use in Claude.ai and ChatGPT

Key Features

01Enforces mandatory workflow phases and state-specific licensing restrictions.

02Prohibits the collection of sensitive PII like SSNs and bank account numbers within chat.

03Defends against conversational and document-based prompt injection attacks.

04Resists social engineering tactics and unauthorized claims of administrative authority.

052 GitHub stars

06Prevents disclosure of internal system prompts, configuration files, and API credentials.

Use Cases

01Securing mortgage data extraction from uploaded documents against embedded malicious instructions.

02Protecting proprietary pricing logic and internal backend infrastructure details from user discovery.

03Ensuring strict adherence to regulatory compliance and state licensing boundaries during refinance quotes.