Skip to main content
affpapa.org
RU EN
subagent

llm-redteam-specialist — Claude Code subagent

Subagent designs adversarial probe suites, runs jailbreak evaluations, and drafts compliance evidence packages for deployed language models.

Install into Claude Code · claude-code-templates All methods and platforms ↓ 
npx claude-code-templates@latest --agent security/llm-redteam-specialist
Open on GitHub

What llm-redteam-specialist does

The subagent structures adversarial assessments for both cloud APIs and local deployments using Ollama, vLLM, or llama.cpp. It covers direct jailbreak families, prompt leaking, context manipulation, and indirect injections embedded in retrieval contexts.

It operates with Read, Grep, Glob, and Bash to inspect repositories and configure test runners. Grading relies on deterministic keyword or regex patterns in restricted enclaves, or calibrated model-as-judge setups when external networks are permitted. Findings map directly to NIST AI RMF MEASURE-2.7, EU AI Act Article 15, and the OWASP LLM Top 10 taxonomy.

Without network access, the agent defaults entirely to offline rubrics and avoids remote scoring APIs. It can integrate references to tooling such as Garak, Promptfoo, and PyRIT.

What it helps with

  • Build an evidence bundle mapped to EU AI Act Article 15 and NIST AI RMF
  • Design an offline prompt injection test harness for Ollama endpoints
  • Audit retrieval pipelines against indirect injection through external documents
  • Generate an adversarial test report aligned with the OWASP LLM Top 10

Example prompts for the agent

  • Build an air-gapped test suite to check our vLLM deployment for indirect prompt injection
  • Generate a prompt injection compliance report mapped to OWASP LLM Top 10 for enterprise security review
  • Audit our retrieval pipeline prompts against system instruction extraction and obfuscation attacks

How to install llm-redteam-specialist

1claude-code-templates

npx claude-code-templates@latest --agent security/llm-redteam-specialist

2Manually: put the files in place

.claude/agents/llm-redteam-specialist.md

For all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention

Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean

Questions about llm-redteam-specialist

How to install llm-redteam-specialist in Claude Code?

Run npx claude-code-templates@latest --agent security/llm-redteam-specialist in your project root. The agent file is saved to .claude/agents/llm-redteam-specialist.md.

What does llm-redteam-specialist do?

It configures adversarial probes, checks output safety, scores failures, and packages audit-ready reports for regulatory reviews.

Can it run in air-gapped environments without external APIs?

Yes, it targets localhost endpoints like Ollama and uses deterministic, regex-based rubrics rather than external cloud graders.

Similar subagents

All subagents for Security →
code-reviewer official Anthropic This subagent checks code changes for bugs, security issues, and project-rule violations. Use it when a review should focus on findings that matter. 31,050installs security-auditor official Anthropic This subagent searches code for security flaws and gives engineers concrete fixes. Use it to review a project before hardening or modernization work. 10,718installs test-engineer official Anthropic This subagent captures legacy behavior in tests and compares it with a rewritten implementation. Use it before changing existing modules. 9,715installs code-reviewer VoltAgent Community Reviews code changes for bugs, security flaws and maintenance issues. Use it before merging complex work or auditing a project. 31,050installs security-auditor VoltAgent Community Audits security controls, vulnerabilities, risks, and compliance gaps across systems, infrastructure, and processes. Use it when you need documented findings. 10,718installs devops-engineer VoltAgent Community This subagent reviews infrastructure and helps with CI/CD, containers, and deployment workflows. Use it when you want to automate software delivery or add security checks to the process. 7,392installs

From the same set804

The whole set →
perplexity This skill wires Perplexity into Claude: web search, conversational answers, and deep research through an agent. Handy when you need current information rather than library docs. 140installs cross-reference-manager Audit and repair references between GitHub and Linear, find orphaned links, and export mapping reports. Useful when you need to check cross-platform link consistency. 140installs api-fuzzing-bug-bounty This skill guides security checks for REST, SOAP, and GraphQL APIs. Use it to structure authorized bug bounty work or an API security assessment. 139installs meme-factory Builds meme image links via the memegen.link service inside Claude Code. Useful for social media graphics, code review jokes, and team chats. 139installs feature-pipeline Executes coding tasks sequentially from a markdown design document and updates checkboxes directly in the file without interruptions. 137installs dependency-updater A skill that detects your project type, applies safe dependency updates on its own and asks before touching major versions. 136installs sync-conflict-resolver Resolves data synchronization discrepancies in files like .sync-state.json. Useful for automated field-level merging and conflict error log reviews. 135installs data-processing-ray-data A Claude skill that tells the agent how to process large datasets with Ray Data: streaming, GPU transforms, batch inference. 134installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor