Individual AI skill · AI for AI Agents and Frameworks

prompt-guard

1200+ pattern AI agent security defense covering prompt injection, indirect injection, multimodal attacks, MCP/tool poisoning, supply chain injection, memory poisoning, RAG poisoning, cross-agent propagation, skill weaponization, ASCII smuggling, unicode steganography, action gate bypass, cascade amplification, multi-turn manipulation, authority escalation, PII/cloud credentials DLP, code exfiltration, guardrails bypass, and agentic workflow exploitation. Platform-agnostic — works on Claude Code, Antigravity/Gemini CLI, Codex, Copilot, Cursor, and any SKILL.md-compatible agent.

Source content · review pending Source language: EN

Choose how you use AI

For projects ChatGPT · Gemini · Claude For agents Cursor · Claude Code · Codex

See how to use it

Free ZIP · One canonical method · Usage example · Installation guide · Licence notices

What you receive

One method. No filler.

01
SKILL.md / Project instructions
The same method, wrapped for your way of working.
02
EJEMPLO-DE-USO.md
A prompt you can adapt to your case.
03
INSTALAR-CON-TU-IA.md
The route for web interfaces and agents.

Version: 1.0.1
Format: Platform-agnostic
Access: Free with account

From file to useful result

How to ask with this skill installed

Installing the skill gives your AI a method. Your request still has to provide the case-specific facts, constraints and expected output.

1Name the real task

Describe the decision or deliverable, not just the topic.

2Provide the evidence

Add source material, audience, limits and known facts.

3Define done

Set format, quality criteria and checks.

Adaptable starting prompt Replace the brackets with your case

Apply the "prompt-guard" skill to this task.

Goal: [describe the result you need]
Context: [company, audience, situation and constraints]
Source material: [paste or attach the relevant information]
Required output: [format, length and acceptance criteria]

Before producing the result, list any essential information that is missing. Do not invent facts. State assumptions and include a final verification checklist.

Choose the right scope

Packs that include this skill

Download only this method or take it together with the rest of its professional area.

This page

Individual skill

The smallest useful download, prepared for a web project or an agent.

Choose format →

Professional pack

AI for AI Agents and Frameworks

This skill plus the other methods selected for this professional area.

Explore the pack →

Same method, different setup

Where will you use it?

We do not duplicate the skill for every platform. We keep one canonical method and explain the correct installation route.

01 ChatGPT

Use it as project knowledge and instructions.

Open web-interface guide → 02 Claude

Add it to a Project and keep the case material separate.

Open web-interface guide → 03 Agents and IDEs

Install SKILL.md in the path expected by your agent.

Compare agent guides →

Traceable by design

Source and licence travel with the file.

SkillsBundle catalogues material of its own and from third parties. Inclusion does not claim original authorship. The individual ZIP preserves the canonical catalogue path and includes the applicable licence notices.

Catalogue path: skills/transversal/prompt-guard
Ficha origin: Extracted from SKILL.md

Read terms and licence policy →

Start with one real task

Install the method. Keep your judgement.

Choose format