LLM App Security

Use when designing or reviewing an LLM-powered application for prompt injection, unsafe outputs, data leakage, tenant isolation, and abuse risks.

9 stars 3 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add sandbaseai/workbuddy-skill --skill llm-app-security
Manual install
git clone --depth 1 https://github.com/sandbaseai/workbuddy-skill.git /tmp/workbuddy-skill
mkdir -p ~/.claude/skills
cp -r /tmp/workbuddy-skill/skills/llm-app-security ~/.claude/skills/llm-app-security

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# LLM Application Security

Review an application that accepts untrusted input, invokes a language model, retrieves data, or calls tools. Treat the model as an untrusted probabilistic component: enforce security decisions in deterministic application and service layers. Work only within an authorized scope; never exfiltrate secrets, bypass access controls, weaponize prompts, or send harmful content to users or external systems.

## Map the trust boundaries

Inventory user input, uploaded files, system and developer instructions, retrieved documents, conversation memory, model providers, tools, plugins, webhooks, logs, analytics, and downstream actions. For each flow, identify the principal, tenant, data classification, authorization decision, side effect, and owner. Keep system instructions separate from user-controlled content, and mark every assumption or unknown.

## Verify the control layers

Inspect code, configuration, prompt templates, retrieval filters, tool schemas, identity middleware, moderation policy, rate limits, telemetry, and tests. Check at minimum:

- **Input and context:** validate type, size, encoding, file content, and provenance; scan uploads for secrets or malware where appropriate; isolate retrieved and user text from instructions;
- **Identity and tenancy:** derive authorization from trusted identity state, filter retrieval and memory by tenant and object permissions, prevent cross-tenant caching, and deny by default when context is ambiguous;
Read the full skill on GitHub

Details

Skill name
llm-app-security
Path
skills/llm-app-security/SKILL.md
License
MIT
Repo updated
8 September 2026

FAQ

What does the LLM App Security skill do?

Use when designing or reviewing an LLM-powered application for prompt injection, unsafe outputs, data leakage, tenant isolation, and abuse risks.

How do I install the LLM App Security skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add sandbaseai/workbuddy-skill --skill llm-app-security. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is LLM App Security popular?

It lives in sandbaseai/workbuddy-skill, which has 9 GitHub stars and 3 forks.

Related skills

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.