CCDV-F Security Practice Question
Your team is building an agentic workflow that interacts with internal databases. Which TWO security practices should be implemented to prevent prompt injection attacks that could lead to unauthorized data exfiltration?
⚠ Common exam trap
Candidates mistakenly rely on the model's safety training alone, forgetting that agentic workflows need strict function allow-lists and delimiter wrapping to prevent malicious exfiltration.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Implement a strict allow-list of tools and functions the model can execute.
Preventing prompt injection requires a defense-in-depth approach that separates user input from system instructions and enforces strict operational boundaries. By limiting the agent's capability to only necessary functions and using structured output formats, developers minimize the attack surface. These practices are critical because agentic workflows are highly susceptible to malicious instructions that override original system prompts, potentially leading to unauthorized data queries or unintended execution of dangerous internal operations.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Use hardcoded system prompts that are strictly enforced via fine-tuning.
Why it's wrong here
While fine-tuning can influence model behavior, it is not a primary defense against prompt injection. An attacker can still provide input that overrides the fine-tuned tendencies. Reliance on fine-tuning for security provides a false sense of safety, as dynamic input processing always risks exploitation without other structural controls.
- ✓
Implement a strict allow-list of tools and functions the model can execute.
Why this is correct
Limiting tool usage to a curated allow-list prevents the agent from calling unauthorized APIs or database functions. Even if an injection attack successfully manipulates the model, the model is unable to trigger unintended actions because the execution environment rejects any non-whitelisted function calls, effectively containing the potential damage.
- ✓
Wrap user input in XML tags or specific delimiters and instruct the model to treat content within tags as untrusted data.
Why this is correct
Using XML tags to explicitly delineate user input from system instructions helps the model distinguish between executable commands and untrusted data. This structural boundary is a core security design pattern for LLM applications, as it provides clear context, significantly reducing the success rate of malicious prompt injection attempts.
- ✗
Perform all API calls using a public-facing read-only database user.
Why it's wrong here
Although using a read-only account limits the scope of data modification, it does not prevent data exfiltration. An attacker could still read all sensitive information from the database and send it to an external endpoint. This control is insufficient for preventing the disclosure of confidential data via prompt injection.
- ✗
Require human-in-the-loop approval for all model responses.
Why it's wrong here
Human-in-the-loop is an operational control rather than a security architecture defense against prompt injection. While useful for quality control, it does not prevent the vulnerability from being triggered in the first place, and it may be bypassed by automated agents processing large volumes of data without human oversight.
About these practice questions
This CCDV-F question is part of Courseiva's 257-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCDV-F practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCDV-F exam.