PURPOSE
What this family supports.
Workflow rails, permission boundaries and contamination-path instruments for tool-using AI and retrieval systems.
TOOL METHODOLOGY
Agentic/RAG Tool Methodology.
Additional page sections
Workflow rails, permission boundaries and contamination-path instruments for tool-using AI and retrieval systems.
INPUT ASSUMPTIONS
How inputs are treated.
Inputs are user-entered assumptions or evidence states. The tools do not verify live systems, datasets, vendors, clinical claims, vulnerabilities or scientific measurements.
OUTPUT POLICY
Artifact boundaries.
Outputs are preliminary orientation artifacts for human review. They are not certification, legal advice, medical advice, engineering sign-off or compliance approval.
OUTPUT ARTIFACTS
Artifacts produced.
- Agentic workflow control map
- Permission-boundary map
- RAG contamination path report
- Prompt-injection defensive test case
- Model/RAG release memo
- AI vendor assurance questionnaire
SOURCE FAMILIES
Reference families.
AI procurement governanceAI release governanceOWASP LLM guidanceRAG security controlsRAG security testingRAG threat modelingagentic AI governanceagentic workflow assuranceaudit logginginformation retrieval assuranceleast privilegeprompt-injection testing practicesecure SDLCsecure supplier reviewsecure workflow designsecure-by-design AI systems
INSTRUMENTS
Tools using this methodology.
Agentic AI and RAG Security
Additional section
Agentic Workflow Rail
Builds a swimlane control map for user, agent, policy, memory, retrieval, tools, human approval, audit and rollback.
- Best for
- Agentic workflow control map
- Input
- 6 categorical evidence fields
- Output
- Agentic workflow control map
- Method
- swimlane
- Maturity
- Beta · v1.9
- Limits
- Does not simulate a real agent.
- Exports
- copy, Markdown, JSON, Mermaid
Agentic AIRAG Security
open instrument →
Agentic AI and RAG Security
Agentic AI Permission Boundary Designer
Creates an allow/deny permission-boundary architecture for tool-using AI systems.
- Best for
- Permission-boundary map
- Input
- 6 categorical evidence fields
- Output
- Permission-boundary map
- Method
- permission graph
- Maturity
- Beta · v2.1
- Limits
- Does not generate bypass instructions.
- Exports
- copy, Markdown, JSON, Mermaid
Agentic AIRAG Security
open instrument →
Agentic AI and RAG Security
RAG Contamination Threat Model
Maps contamination paths from source material through retrieval into model output and downstream action.
- Best for
- RAG contamination path report
- Input
- 6 categorical evidence fields
- Output
- RAG contamination path report
- Method
- pipeline
- Maturity
- Beta · v2.3
- Limits
- Does not test a live vector database.
- Exports
- copy, Markdown, JSON, Mermaid
Agentic AIRAG Security
open instrument →
Agentic AI and RAG Security
Prompt-Injection Sandbox
Classifies suspicious instruction patterns in test text and maps them to defensive containment recommendations.
- Best for
- Prompt-injection defensive test case
- Input
- 4 structured fields including narrative context
- Output
- Prompt-injection defensive test case
- Method
- console lab
- Maturity
- Released · v2.4
- Limits
- Does not provide bypass recipes.
- Exports
- copy, Markdown, JSON
Agentic AIRAG Security
open instrument →
Agentic AI and RAG Security
Model and RAG Release Gate Checklist
Produces a go/no-go release memo for models, RAG systems, copilots and agentic features.
- Best for
- Model/RAG release memo
- Input
- 7 categorical evidence fields
- Output
- Model/RAG release memo
- Method
- timeline
- Maturity
- Beta · v2.5
- Limits
- Does not approve release.
- Exports
- copy, Markdown, JSON
Agentic AIRAG Security
open instrument →
Agentic AI and RAG Security
AI Vendor Assurance Questionnaire Builder
Generates buyer or vendor assurance questions for model APIs, copilots, RAG platforms and agentic workflow vendors.
- Best for
- AI vendor assurance questionnaire
- Input
- 5 categorical evidence fields
- Output
- AI vendor assurance questionnaire
- Method
- evidence matrix
- Maturity
- Beta · v2.7
- Limits
- Does not verify vendor claims.
- Exports
- copy, Markdown, JSON, CSV
Agentic AIRAG Security
open instrument →
Version policy: each instrument has a version, maturity state, assumptions, limitations, example input, example output and export formats. Method notes should be updated when scoring logic, input taxonomy or source families change.