Assessment Term Sheet¶
Use this plain-language sheet while reading the entry diagnostic, exit diagnostic, or chapter checkpoints. Looking up a term is allowed and does not reveal a recommended decision, score, or answer pattern.
| Term | Plain-language meaning |
|---|---|
| Access token | A secret credential that lets software act on an account. A broad token grants many permissions or reaches a large scope of data. |
| Artifact | A concrete piece of work used as evidence, such as a file, screenshot, recording, test result, or deployed page. |
| Authorization | Permission from the person or organization responsible for data or a system. |
| Data classification | A label for how sensitive data is and what handling rules apply to it. |
| Data minimization | Using, sharing, and retaining only the data necessary for a specific task. |
| Fixture | A known input or saved setup used to run a repeatable test. |
| Least privilege | Giving a person or tool only the access needed for the current task, for only as long as it is needed. |
| MCP server | A program that offers tools or data to an AI agent through the Model Context Protocol. |
| Monitoring | Repeatedly observing a released system to detect failures, misuse, cost changes, or unexpected behavior. |
| Prompt injection | Instructions hidden or embedded in content that try to redirect an AI system or its tools. |
| Provenance | A record of where information or an artifact came from and how it was produced or changed. |
| Provider handling | How an AI or cloud provider sends, stores, retains, reviews, and uses submitted data. |
| Recourse | A way for an affected person to question, correct, appeal, or obtain human review of an output or decision. |
| Regression | A behavior that used to work but breaks after a change. |
| Revocation | Removing access that was previously granted, such as disabling a credential or disconnecting a tool. |
| Roll-forward | Recovering by making and deploying a new corrective change instead of restoring an older version. |
| Rubric | A scoring guide that names the criteria and evidence used to judge work. |
| Sandbox | An enforced environment that limits which files, processes, networks, or other resources a program can reach. |
| Schema | A description of a data structure: its fields, types, relationships, and rules. |
| Synthetic data | Artificial examples that resemble a useful pattern without containing real confidential or personal information. |
These definitions explain vocabulary only. Your assessment response must still decide which risks matter, what evidence is sufficient, and what action is justified in the situation.