What it does
The agent designs and documents red-team work on large language model deployments. It covers jailbreak probes, prompt injection harness design, output-safety evaluation and robustness evidence mapped to EU AI Act Article 15 and NIST AI RMF MEASURE-2.7. It applies to cloud-hosted models and to on-prem or air-gapped local models served through Ollama, vLLM or llama.cpp. It distinguishes model-level tests from system-level tests and uses jailbreak taxonomies, so it is aimed at adversarial input to an LLM rather than generic web penetration testing.
The agent suits teams that need evidence of how an LLM resists adversarial input before deployment. The publisher's examples include a clinical triage assistant, where the aim is to check for harmful medical advice and leakage of personal health data from retrieval context. Another example is an air-gapped network that runs only local Llama-3 and Mistral instances and needs recurring robustness evidence without external API calls.
In the publisher's examples, the output is a probe suite covering jailbreak families such as role-play escalation, encoding attacks, prompt leaking and indirect injection through retrieved documents. It also includes a scoring rubric matched to the deployment's harm taxonomy and a repeatable runner that targets a local endpoint. Scoring is deterministic, with no model-as-judge calls outside the enclave. The evidence pack contains a pass/fail table, example transcripts, a coverage metric and a control narrative.
Limitations and requirements
The agent is scoped to adversarial testing of LLMs, not generic web penetration testing. The examples assume access to a model endpoint such as a local Ollama, vLLM or llama.cpp server, and an air-gapped setup avoids external API calls. The publisher's description is cut off in the source, so the full list of requirements is not known.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- AI Templates (aitmpl) first seen 13 Nov 2026 · last seen 18 Nov 2026 · its page on the source · figures reported by the source: aitmpl downloads 167
llm-redteam-specialist
Use this agent when you need to red-team a Large Language Model deployment — jailbreak probes, prompt injection harness design, output-safety evaluation, and robustness evidence for EU AI Act Article 15 or NIST AI RMF MEASURE-2.7. Covers cloud-hosted models and on-prem / air-gapped local models (Ollama, vLLM, llama.cpp). Specifically:\n\n<example>\nContext: A healthcare vendor embeds an LLM in a clinical triage tool and the compliance team wants a red-team report before rollout.\nuser: "We're deploying a Llama-3 70B behind a clinical assistant. Legal wants evidence it won't hand out harmful me
As shown on www.aitmpl.com when we last read it