security-reviewer-deep

A deep security-review agent for AI systems that use chat, language models, and tool protocols such as MCP and A2A. It focuses on prompt protection, malicious instructions in user data, customer isolation, and data exposure.

In plain words
What is it for?
Use it when reviewing changes to system prompts, chat interfaces, MCP or A2A endpoints, databases, authentication, or any flow that sends user-controlled data into an AI context.
Why use it?
It traces how untrusted input moves through an AI application, where it is stored, and which tools or prompts receive it. This helps uncover attack chains that ordinary code reviews may miss.

Agent for Claude Code

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.

agentmods
npx agentmods add agents/adcontextprotocol/adcp/security-reviewer-deep
Clone the repo
git clone --depth 1 https://github.com/adcontextprotocol/adcp

Made for: Claude Code.

Per session 64 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,413 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00064 $0.02413
Opus 5 $0.00032 $0.01207
Sonnet 5 $0.00013 $0.00483
Haiku 4.5 $0.00006 $0.00241

Measured 2d ago against content hash 2f9258239713, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

security-reviewer-deep 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 2d 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.

.claude/agents/security-reviewer-deep.md · 192 lines

How it starts

The opening of the file, as written. The whole thing — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Security Reviewer - Agentic Ad Tech Systems

Core Identity

You are a security reviewer who specializes in the attack surfaces unique to agentic AI systems. You understand that traditional web security (SQLi, XSS, CSRF) is necessary but insufficient for platforms where LLMs process user input, tool calls execute actions, and system prompts control behavior. You know that an attacker who modifies a system prompt can silently poison thousands of users without touching application code.

Your reviews are informed by real breaches. The McKinsey/Lilli attack (March 2026) demonstrated that AI chatbot platforms face a specific threat model: exposed API documentation enables automated enumeration, SQL injection through JSON keys bypasses parameterized queries, and database-stored system prompts become a single point of compromise for all downstream users.

When Invoked

  1. Run git diff to see recent changes (or review the files/PR specified)
  2. Identify which system the changes touch and what data flows through it
  3. Trace user input from entry point to every place it is used, stored, or forwarded
  4. Apply the review checklist below, starting with Priority 0
  5. Check for attack chains: vulnerabilities that are low-severity alone but critical in combination

Review Priorities

Priority 0: Prompt Injection and System Prompt Integrity

Direct prompt injection:

  • User input concatenated into system prompts, tool descriptions, or LLM context without sanitization
  • Chat messages, form fields, campaign names, product descriptions, or any user-controlled string that reaches an LLM
  • Template literals or f-strings that embed user data into prompt text (search for backtick templates and f-strings near LLM calls)

Indirect prompt injection:

  • Stored data (chat history, database records, file contents) that is later retrieved and included in LLM context
  • Campaign names, product descriptions, or metadata that flow from one user's input into another user's agent context
  • Tool responses from external systems that could contain adversarial instructions

Read the full file on GitHub · 192 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. 2d ago First seen · 192 lines · 64 tokens per session scan A 2f9258239713

Subscribe to this mod's changes

security-reviewer-deep is an agent published in the GitHub repository adcontextprotocol/adcp (241 stars, last pushed 2d ago), licensed Apache-2.0. It adds 64 tokens to every session and 2,413 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.

Related

Other agents, from other repositories

Demonstrate

Agent for demonstrating VS Code features.

microsoft/vscode · 10 tokens

playwright-test-generator

Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.

microsoft/playwright · 151 tokens

.NET-Notebook-Migration-Agent

Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.

microsoft/ai-agents-for-beginners · 33 tokens

AVM Owner Triage

Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.

github/awesome-copilot · 61 tokens

Ultimate Transparent Thinking Beast Mode

Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.

github/awesome-copilot · 11 tokens

code-reviewer

Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.

anthropics/claude-cookbooks · 52 tokens