Prompt Injection Defense

Use when threat-modeling or hardening an AI agent, RAG system, assistant, or tool workflow against direct, indirect, stored, cross-agent, or multimodal prompt injection.

9 stars 3 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add sandbaseai/workbuddy-skill --skill prompt-injection-defense
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/prompt-injection-defense ~/.claude/skills/prompt-injection-defense

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

From the SKILL.md

# Prompt Injection Defense

Design for compromise of model reasoning. Prompt text, delimiters, classifiers,
and detectors are defense-in-depth signals, not reliable security boundaries.
Keep consequential authority, authorization, validation, and policy enforcement
outside the model. Never test against production, third-party, or user systems
without target-specific authorization.

## Map the attack surface

Collect the agent purpose, instructions, models, memory, orchestration, every
input source and modality, tools and privileges, identities and targets,
secrets and protected assets, output sinks, egress, authorization model,
human gates, monitoring, incident history, and safe evaluation environment.
Include web pages, email, documents, images, audio, OCR, metadata, tool
descriptions, retrieval results, persisted memory, and other agents. Label
assumptions and unknowns. Use redacted samples or synthetic fixtures; never
request production secrets or malicious artifacts in chat.

Produce a data-flow and trust-boundary map, asset/consequence inventory,
injection-path threat model, prioritized controls, regression plan, evidence
report, and residual-risk/recovery plan. Track each control as `missing`,
`planned`, `implemented`, or `verified`, with an enforcement point, owner,
evidence ID, test ID, and expiry when time-limited.

## Enforceable boundaries

- Retrieved content cannot grant permissions or change a tool allowlist.
Read the full skill on GitHub

Details

Skill name
prompt-injection-defense
Path
skills/prompt-injection-defense/SKILL.md
License
MIT
Repo updated
8 September 2026

FAQ

What does the Prompt Injection Defense skill do?

Use when threat-modeling or hardening an AI agent, RAG system, assistant, or tool workflow against direct, indirect, stored, cross-agent, or multimodal prompt injection.

How do I install the Prompt Injection Defense skill?

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

Is Prompt Injection Defense popular?

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

Related skills

Access Control Security Best Practices

9

sandbaseai/workbuddy-skill

Design, review, implement, and verify least-privilege access controls across users, services, tenants, resources, and administrative workflows.

Security

Accessibility Review

9

sandbaseai/workbuddy-skill

Use when reviewing a web interface or interaction for semantic, keyboard, focus, screen-reader, ARIA, and contrast accessibility risks.

Frontend & Design

Acquire Codebase Knowledge

9

sandbaseai/workbuddy-skill

Use when the user asks to map, document, or onboard into an existing codebase and needs evidence-backed architecture, structure, conventions, integrations, testing, and concerns documentation.

Software Development

Prompt Optimizer

9

sandbaseai/workbuddy-skill

Turn any rough prompt, half-formed idea, or task description into a finished, ready-to-send prompt optimized for any LLM model inside a chat interface — NOT the API. Use this skill whenever the user wants to write, rewrite, optimize, improve, sharpen, or polish a prompt for chat. Trigger phrases include "rewrite this prompt", "make this a better prompt", "optimize this prompt", "turn this into a prompt", "help me prompt this", "draft a prompt that...", "I want to ask...", or whenever the user pastes a draft prompt and asks for improvements. Also trigger when the user describes a task they plan

AI, LLMs & Agents

Prd

9

sandbaseai/workbuddy-skill

Use when turning a product or AI feature idea into a measurable Product Requirements Document with user stories, technical specifications, acceptance criteria, evaluation, risks, and rollout phases.

AI, LLMs & Agents

RAG Evaluation

9

sandbaseai/workbuddy-skill

Use when designing, running, reviewing, or diagnosing evaluations for retrieval-augmented generation systems, including corpora, chunking, retrieval, reranking, context assembly, grounded answers, citations, refusals, latency, cost, and regressions.

AI, LLMs & Agents

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