@orion-blakesecurity-risk-rag-context-packTextÖffentlichAktualisiert am 14.06.2026

Security Risk prompt that turns source material into a reliable retrieval design and returns source policy, chunking plan, retrieval prompt, citation rules, and freshness checks.

AI Safety AI Workflow Privacy Risk Security

38Stars0Fork219Kopien

Prompt

Vorschau

In Claude öffnen

# Role
You are a pragmatic security architect who finds exploitable risk without blocking useful shipping. You know current AI product patterns: tool calling, structured outputs, long-context reasoning, multimodal generation, MCP connectors, evals, trace review, and human approval gates.

# Mission
Turn source material into a reliable retrieval design for AI system threat modeling, prompt-injection review, data exposure risk, and incident readiness.

# Inputs
- Project context: an internal agent that can read tickets, GitHub issues, and customer documents
- Target audience: security engineers, platform owners, privacy teams
- Success metric: activation, quality, revenue, risk reduction, or cycle time
- Available tools and data: threat model template, SIEM, secret scanner, policy engine
- Constraints: prioritize exploitability; map each risk to a control; respect privacy boundaries
- Desired depth: Fast triage
- Output tone: Clear operator memo

# Domain signals to inspect
- data flow diagrams
- tool permissions
- audit logs
- security incidents

# Known risks to control
- prompt injection
- credential leakage
- cross-tenant access
- unreviewed tool writes

# Method
1. Classify sources by authority and freshness
2. Design chunk boundaries around user intent
3. Require citations for claims
4. Create known-answer and adversarial queries

# Required output sections
1. Source inventory
2. Chunking strategy
3. Retrieval query plan
4. Citation contract
5. Failure handling
6. Evaluation set

# Output contract
Return Markdown only.
Use concise headings, decision tables, and checklists.
Every recommendation must tie back to evidence, assumptions, or a missing-input note.

# Quality bar
- Make the output reusable by another practitioner without extra explanation.
- Separate facts, assumptions, and recommendations.
- Use variables and placeholders only where the user must fill in project-specific information.
- Ask at most 3 clarifying questions only if missing information would materially change the result.
- Do not invent private facts, metrics, laws, integrations, or user quotes.
- Prefer specific operational language over generic AI enthusiasm.
- Include a "Verification" subsection when the output could affect users, money, security, compliance, health, or public trust.
- If the task involves tools or automation, state which actions are read-only, which actions write data, and which actions require human approval.

Artefakte

1 Artefakte

Example Output: Security Risk RAG Context Pack

Inputs used

Project context: an internal agent that can read tickets, GitHub issues, and customer documents
Target audience: security engineers, platform owners, privacy teams
Success metric: activation, quality, and risk reduction
Available tools and data: threat model template, SIEM, secret scanner, policy engine
Desired depth: Production-ready
Output tone: Clear operator memo

Generated Result

source policy, chunking plan, retrieval prompt, citation rules, and freshness checks

Source inventory

Rank sources by authority before retrieval. Chunk around task intent rather than page boundaries, and require every answer to cite the exact source segment used for data flow diagrams.

Chunking strategy

Rank sources by authority before retrieval. Chunk around task intent rather than page boundaries, and require every answer to cite the exact source segment used for tool permissions.

Retrieval query plan

Create at least 12 golden tasks: 6 normal cases, 3 edge cases, and 3 adversarial cases targeting cross-tenant access. A passing result must cite the evidence source and state confidence.

Citation contract

Rank sources by authority before retrieval. Chunk around task intent rather than page boundaries, and require every answer to cite the exact source segment used for security incidents.

Failure handling

Use data flow diagrams as evidence, apply the constraint "map each risk to a control", and explicitly note how the plan reduces prompt injection. The output should be ready for a practitioner to act on without a follow-up explanation.

Evaluation set

Create at least 12 golden tasks: 6 normal cases, 3 edge cases, and 3 adversarial cases targeting credential leakage. A passing result must cite the evidence source and state confidence.

Recommended Decision

Proceed with a narrow pilot focused on data flow diagrams and tool permissions. Treat prompt injection as the primary launch blocker. The first milestone should prove that the workflow produces a usable risk register, mitigations, and verification checklist with clear evidence, named owners, and a review path for ambiguous cases.

Expected quality checks

The result is specific to AI system threat modeling, prompt-injection review, data exposure risk, and incident readiness.
It includes the required sections: Source inventory, Chunking strategy, Retrieval query plan, Citation contract, Failure handling, Evaluation set.
It separates evidence, assumptions, risks, and recommended next actions.
It includes practical verification steps, not only generic advice.
It names the most important failure mode for this domain: prompt injection.

Reuse note

Before copying the output into production work, replace all default variables with your real data and run a human review for high-impact decisions.

README

README.md

Security Risk: RAG Context Pack

Use this prompt when you need source policy, chunking plan, retrieval prompt, citation rules, and freshness checks for AI system threat modeling, prompt-injection review, data exposure risk, and incident readiness.

Best for

security engineers, platform owners, privacy teams
Teams that already have partial context but need a sharper, reusable artifact
AI workflows where the output must be auditable, editable, and easy to hand off

How to use

Replace the variables in the prompt with your real project context.
Keep the default constraints unless your team has stronger internal rules.
Review the generated output against the checklist in the example artifact.

Design notes

This seed follows current prompting practice: explicit role, structured inputs, domain evidence, operational guardrails, and a concrete output contract. It is written in English for international PromptHub users.