CCA-F Exam Prep
CCA-F Exam Glossary — 30 Terms
Search the CCA-F terminology pack covering Claude Code, agentic architecture, MCP tool design, structured output, context windows, and production reliability patterns.
A
- API key
- A credential used to authenticate access to an API or service.
C
- CLAUDE.md
- A persistent instruction file that provides guidance to Claude Code sessions but does not enforce runtime actions.
- context window
- The bounded amount of information a model can access and process in a single interaction.
- conversation history
- The accumulated prior messages and interactions available to an agent as context.
- coordinator-subagent pattern
- An orchestration design where a coordinator delegates subtasks to subagents and combines their results.
- custom slash command
- A reusable command invoked with a slash prefix to package multi-step workflows into one action.
E
- errorCategory
- A structured field classifying the type of error returned by a tool or system.
- ETL
- Extract, transform, load; a sequential data-processing pipeline where each stage depends on the previous one.
H
- hook
- An execution-time interception mechanism that can enforce, modify, or block actions.
I
- intermediate context
- Temporary working information generated during subtask processing before final aggregation.
- isError
- A response field indicating whether a tool call resulted in an error.
L
- latency
- The time delay between initiating a request and receiving its result.
- lifecycle events
- Named stages in Claude Code execution, such as session start, prompt submission, tool use, and stop.
M
- MCP tool
- A tool integrated through the Model Context Protocol that returns structured responses to agent requests.
- model calls
- Requests made to an AI model for inference or generation.
O
- overhead
- Additional computational or operational cost introduced by a design choice without proportional benefit.
P
- parallelism
- The ability to perform multiple tasks concurrently to reduce overall latency.
R
- refresh policies
- Rules that determine when stored data should be updated to maintain accuracy.
S
- schema validation
- The process of checking whether data conforms to a declared structural schema.
- shell command
- A command executed in a system shell to perform operating system or developer tasks.
- stale-data detection
- A method for identifying stored information that may no longer be current or reliable.
- strict JSON
- Output constrained to valid JSON structure suitable for reliable downstream parsing.
- structured error context
- A detailed error payload containing failure type, failed input, partial results, and recovery-relevant metadata.
- subagent context
- The separate context window assigned to a subagent, distinct from the parent agent’s conversation history.
- subagent invocation
- A single execution or call that launches a subagent to perform a task.
T
- tokens
- Units of text processed by a language model for input and output accounting.
- tool actions
- Operations performed through integrated tools during agent execution.
- tool selection
- The model’s process of choosing the most appropriate registered tool for a task.
U
- usage-based costs
- Costs incurred based on actual model or tool usage, such as tokens, model calls, or tool actions.
W
- workflow
- A defined sequence of steps or operations executed to complete a task.
About These Definitions
These definitions are loaded from the CCA-F terminology release pack in ExamPal shared content. Use them with the study guide and practice questions to connect vocabulary to architecture decisions and tool-use scenarios.
Practice with 490+ CCA-F questions
ExamPal turns these concepts into scenario-based questions with explanations across all five domains.
Download on App Store