Claude Skill MartBrowse skillsWhat is a Skill?
 
← Back to list

AI Security Assessment (ai-security)

Assess LLM and ML systems for prompt injection, jailbreaks, model inversion, data poisoning and agent tool abuse, mapped to MITRE ATLAS.

Security & ReviewAdvanced★ 24,151⑂ 3,405AI score 8/10Last updated: Aug 9, 2026

What it does

This is not general appsec — it targets AI/ML systems and LLM agents specifically.

  • Detects prompt injection signatures (role override, indirect injection, system prompt extraction, tool abuse, poisoning markers)
  • Provides a jailbreak taxonomy and resistance-testing procedure
  • Scores model inversion risk by access level (black/gray/white-box)
  • Scores data poisoning risk by fine-tuning scope (fine-tune, RLHF, RAG, inference-only)
  • Maps findings to MITRE ATLAS techniques (AML.T0051, T0056, T0020, T0043, T0024)
  • Guardrail design patterns for input, output and agent layers, plus 7 anti-patterns
  • Ships a CLI (scripts/ai_threat_scanner.py) with exit codes usable as a deployment gate

Who it's for

  • Teams shipping user-facing LLM features who need a pre-launch security review
  • Engineers running RAG or tool-calling agents worried about indirect injection
  • AI red teamers, AppSec engineers and security compliance owners
  • Platform teams adding adversarial prompt testing to CI/CD

Examples

  1. 20-minute pre-launch scan: run built-in seed prompts against a black-box LLM profile, check overall_risk, and block release on critical findings.
  2. RAG hardening review: audit indirect injection vectors in retrieved documents and design content validation before indexing.
  3. CI security gate: scan tests/adversarial_prompts.json, emit a JSON report, and fail the pipeline when overall risk is critical.

· · · Install guide · · ·

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the .gemini/skills/ai-security folder from the GitHub repo alirezarezvani/claude-skills into my ~/.claude/skills/ai-security/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/alirezarezvani/claude-skills.git /tmp/claude-skills && mkdir -p ~/.claude/skills && cp -r /tmp/claude-skills/.gemini/skills/ai-security ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal.
  2. Clone the repo: git clone https://github.com/alirezarezvani/claude-skills.git
  3. Create the skills folder: mkdir -p ~/.claude/skills
  4. Copy the skill: cp -r claude-skills/.gemini/skills/ai-security ~/.claude/skills/
  5. Verify with ls ~/.claude/skills/ai-security and confirm scripts/ai_threat_scanner.py and references/atlas-coverage.md came along (search the repo if missing).
  6. Restart Claude Code and try: "Assess my LLM feature for prompt injection risk."
  7. Only run gray-box/white-box modes with the --authorized flag after you have written authorization to test the target model.
View source on GitHub ↗License: MIT