ai-security

ai-security is a skill for Claude Code from Kaademos/secure-sdlc-agents. It costs 50 tokens per session (1,807 once invoked), scanned A, original, MIT.

A security review method for features that use large language models, such as chatbots, retrieval systems, or AI agents with access to tools. It treats model responses as untrusted data.

In plain words
What is it for?
Mapping the attack surface of AI features, reviewing user input and model output, assessing retrieval and embedding systems, and checking AI agents that can affect files, web services, or databases.
Why use it?
It helps identify risks such as prompt injection, excessive tool access, unsafe output handling, and problems in the systems that supply information to the model.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the secure-sdlc-agents plugin — 4 skills, 8 agents shipped together

Good fit Mapping the attack surface of AI features, reviewing user input and model output, assessing retrieval and embedding systems, and checking AI agents that can affect files, web services, or databases.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/kaademos/secure-sdlc-agents/ai-security
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add Kaademos/secure-sdlc-agents --skill ai-security
Clone the repo
git clone --depth 1 https://github.com/Kaademos/secure-sdlc-agents

Made for: Claude Code.

Or install secure-sdlc-agents, the plugin that ships this one along with the rest of its 4 skills, 8 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for ai-security

README.md
[![agentmods](https://agentmods.dev/badge/skills/kaademos/secure-sdlc-agents/ai-security.svg)](https://agentmods.dev/skills/kaademos/secure-sdlc-agents/ai-security)
Your own site
<a href="https://agentmods.dev/skills/kaademos/secure-sdlc-agents/ai-security"><img src="https://agentmods.dev/badge/skills/kaademos/secure-sdlc-agents/ai-security.svg" alt="Measured on agentmods" height="20"></a>
Per session 50 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,807 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00050 $0.01807
Opus 5 $0.00025 $0.00903
Sonnet 5 $0.00010 $0.00361
Haiku 4.5 $0.00005 $0.00181

Measured 8d ago against content hash 853dc5cd5473, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

ai-security scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/ai-security/SKILL.md · 164 lines

How it starts

The opening of the file, as written. The whole thing — 164 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AI Security

Overview

This skill applies structured security analysis to AI and LLM-powered features. The threat categories here — prompt injection, excessive agency, output misuse, supply chain — did not exist before 2023 and are still being misunderstood by most developers shipping AI features today.

Working assumption: every model is a trust boundary, not a trusted component. Model outputs must be treated as untrusted user input to every downstream system.

Reference framework: OWASP Top 10 for LLMs 2025 (LLM01–LLM10).

When to Use

  • Any code that calls an LLM API (OpenAI, Anthropic, Google, Mistral, self-hosted)
  • Any feature that sends user-supplied content to a model
  • RAG systems, embeddings, vector databases, or retrieval pipelines
  • AI agents with tool access (file system, HTTP requests, database writes, email)
  • Features where model output is rendered in UI, executed as code, or used in queries
  • Selecting or integrating a third-party model, fine-tune, or embedding

Process

Step 1 — Map the attack surface

Before finding vulnerabilities, enumerate:

Question Why it matters
Who sends input to the model? Determines direct injection risk
What external sources feed the prompt context? Determines indirect injection risk
What tools / functions can the model invoke? Determines excessive agency blast radius
What happens to the model's output? Determines output handling risk
Is user PII sent to a third-party API? Determines data leakage and legal risk
Where does the model or its weights come from? Determines supply chain risk

Step 2 — Assess prompt injection risk (LLM01, LLM07)

Input trust classification:

Input Source Trust Level Injection Risk
Authenticated user (UI) LOW Direct prompt injection
Public / unauthenticated user UNTRUSTED Direct + jailbreak attempts
Retrieved document (RAG) UNTRUSTED Indirect prompt injection
Tool / function call result MEDIUM Injection via external API response
Database query result MEDIUM Injection via poisoned records
Web scraping / search UNTRUSTED Indirect injection

Read the full file on GitHub · 164 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 164 lines · 50 tokens per session scan A 853dc5cd5473

Subscribe to this mod's changes

ai-security is a skill published in the GitHub repository Kaademos/secure-sdlc-agents (13 stars, last pushed 1mo ago), licensed MIT. It adds 50 tokens to every session and 1,807 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.