Practice — CCAO-F Domain 5: Configuration and Knowledge Management (12%)
12 scenario-based MCQs. Answer key + explanations at the bottom.
Q1
An operations analyst starts each morning by pasting the same 40-page SOP manual and the same formatting rules into a fresh conversation before asking Claude to draft shift summaries. Which configuration best eliminates the repeated setup?
A. Keep one long-running conversation open so the manual never leaves the context B. Add the manual and the formatting rules to global instructions so they apply everywhere C. Create a Project with the manual as a knowledge file and the rules as project instructions D. Save the full paste as a reusable text snippet and insert it at the top of each new chat
Q2
A product manager wants two things: Claude should use plain language and end every deliverable with a bulleted summary no matter what she is working on, and it should apply her team's spec template only when she is working on the "Roadmap" initiative. Which setup matches each requirement to the right layer?
A. Tone and summary rules in global instructions; the spec template in the Roadmap project B. Everything in global instructions so that no requirement is ever missed on any task C. Everything in the Roadmap project so that all of her configuration lives in one place D. Tone and summary rules in the Roadmap project; the spec template in global instructions
Q3
A marketing team uploads its entire archive of 300 past blog posts into a Project's knowledge so Claude "knows the brand." Since then, Claude often ignores parts of the project instructions and quality on long tasks has dropped. What is the most likely cause?
A. The model tier is too small for brand work and should be switched to a larger model B. Project knowledge counts against the context window, and the uncurated archive is crowding it C. Archived posts must be attached through a connector rather than uploaded as knowledge files D. The project instructions are written too briefly to compete with the uploaded documents
Q4
Over a two-hour conversation, a consultant teaches Claude her client's terminology, reporting preferences, and abbreviation glossary. The next day she opens a new conversation and Claude knows none of it. What best explains the situation, and what should she do?
A. Reopen the old conversation — its context syncs into new chats after a short delay B. Remind Claude of the client's name — context from past chats is recalled once re-cued C. Keep all future client work inside the original conversation so nothing is ever lost D. Move the durable material into a Project — instructions for preferences, a knowledge file for the glossary
Q5
A finance team runs the same multi-step "monthly variance report" procedure — same data checks, same structure, same phrasing conventions — but the work happens across several different projects. Team members are tired of re-explaining the procedure. Which mechanism fits best?
A. Paste the procedure document at the start of each report conversation B. Create a skill encoding the procedure, invoked by name or automatically when relevant C. Add the full procedure to each team member's global instructions D. Merge all the projects into one project that holds the procedure document
Q6
One operations specialist has refined a set of skills, reference context, and connector setups that make Claude excellent at customer-email triage. Leadership wants this to become the default workflow for the entire support team. What is the right vehicle?
A. Package the skills, context, and connections into a plugin the team installs B. Share a transcript of the specialist's best session for teammates to imitate C. Have each teammate rebuild the skills by hand from the specialist's notes D. Copy the specialist's setup into global instructions on each teammate's account
Q7
For months, a team member's Claude has appended a legal disclaimer to everything it produces — in every project and in plain chats alike, including brainstorm notes and internal memos where it makes no sense. The disclaimer text is also outdated, and nobody remembers configuring it. Where should they look first?
A. The knowledge files of the currently open project, which may contain the old disclaimer B. The model version, since regressions can resurface behaviors from earlier releases C. Their global instructions, since stale standing context silently distorts every task D. Conversation history, since Claude may be copying the disclaimer from earlier chats
Q8
A sales team's price list lives in a Google Drive spreadsheet that is updated weekly. They uploaded a PDF export as project knowledge, but quotes generated by Claude keep citing superseded prices. Which change best fixes the root cause?
A. Re-export and re-upload the PDF each Monday as part of a team checklist B. Configure the Google Drive connector so Claude reads the live spreadsheet C. Ask a developer to script an API pipeline that regenerates the knowledge file D. Paste the current prices into the conversation each time a quote is needed
Q9
An office coordinator is enabling the Gmail connector so Claude can help manage the team inbox. Following recommended practice, how should access be set up initially?
A. Enable every available connector at the same time so future tasks are never blocked by missing access B. Store the shared inbox credentials inside project knowledge so Claude can authenticate when needed C. Turn off confirmation prompts up front so routine mail-handling tasks don't stall waiting for approval D. Grant access explicitly for this surface and begin with reversible tasks like drafting before sending
Q10
A team packaged a newly written "quarterly report formatting" skill into a plugin and rolled it out. Within a week, every team member's reports contain the same table-formatting error. What practice was skipped?
A. Validating the skill on representative real tasks before packaging it into the plugin B. Restricting plugin installation so that only managers can distribute them team-wide C. Duplicating the formatting rules into each member's global instructions as a backstop D. Splitting the plugin into one plugin per team member to limit the blast radius
Q11
An inventory team wants Claude to answer questions from a proprietary warehouse database that updates hourly. No existing connector supports it. As a Claude associate (not a developer), what is the correct move?
A. Export the database to CSV every hour and upload it as project knowledge B. Paste the relevant database rows into each conversation as questions come up C. Scope the requirement and hand it off to the developer team to build a custom connector D. Add the database's connection string to project instructions so Claude can query it
Q12
A content lead is setting up a "Brand Voice" project. She has a 25-page style guide document and one standing rule: "every draft must be checked against the style guide before delivery." Where does each belong?
A. Both go into project instructions so that Claude never overlooks either of them B. Both go in as knowledge files so that the project instructions stay uncluttered C. The style guide goes in project instructions; the standing rule as a knowledge file D. The standing rule goes in project instructions; the style guide as a knowledge file
Answers
Q1: C. A Project is exactly this: a persistent system prompt (project instructions) plus uploaded knowledge files, built for recurring tasks that need the same reference material. One endless conversation (A) eventually exhausts context and degrades. Global instructions (B) are the wrong scope — they are cross-cutting preferences applied to all work, and 40 pages of task-specific material there would distort unrelated tasks. A saved snippet (D) still repeats the toil and re-spends context every day.
Q2: A. Global instructions hold preferences that apply across all work (tone, formats); a project holds instructions and materials scoped to one initiative. Putting everything global (B) forces the spec template onto unrelated work. Putting everything in the project (C) means her tone rules vanish the moment she works outside Roadmap. (D) inverts the two scopes.
Q3: B. Project knowledge counts against the context window, so an uncurated dump crowds out the instructions and the working conversation — the documented fix is to curate, not upload everything. Switching models (A) is the classic wrong-layer distractor: the issue is context, not capability. Connectors (C) are for live data, not a requirement for archives. Instruction brevity (D) is not the mechanism — volume of knowledge is.
Q4: D. Conversation context lives inside that conversation; it does not automatically carry into new chats. The durable fix is to persist: preferences into project instructions, the glossary into a knowledge file, so every future conversation in the Project starts pre-briefed. (A) and (B) describe cross-conversation recall that doesn't exist here. (C) is the tempting runner-up but loses to context exhaustion as the single conversation grows.
Q5: B. A skill is a reusable, named capability bundle that teaches Claude how the team does one specific task and can be invoked by name or automatically — exactly right for a repeatable procedure needed across multiple projects. Pasting each time (A) is the toil they're trying to remove. Global instructions (C) would push report-specific rules into all unrelated work. Merging projects (D) destroys per-initiative scoping to solve a problem skills already solve.
Q6: A. A plugin is a packaged collection of skills, context, and connections that encodes one expert's workflow into a shareable, installable unit — turning it into everyone's default. A transcript (B) shares an example, not a capability. Manual rebuilds (C) drift and don't scale. Global instructions (D) hold preferences, not a bundled workflow with skills and connections.
Q7: C. Behavior that appears in every task across every project points at the cross-cutting layer: global instructions. The documented risk is exactly this — stale standing context silently distorts all work, so it should be reviewed and pruned like code. Project knowledge (A) is the runner-up but is scoped to one project, and the symptom spans everything. A model regression (B) and copied history (D) don't explain months of consistent, configured-sounding behavior.
Q8: B. The root cause is grounding Claude in a static snapshot of data that changes weekly; the Google Drive connector grounds Claude in the live organizational source. Weekly re-uploads (A) are the plausible runner-up but keep the manual toil and still leave a staleness window between exports. A custom API pipeline (C) is the wrong layer when a standard connector already covers the need. Pasting per quote (D) is repeated toil with the same staleness risk.
Q9: D. Connectors follow a permission-first pattern — access is granted explicitly, per surface — and the best practice is to start with reversible tasks (drafting) before allowing irreversible actions (sending). Enabling everything (A) violates explicit, per-surface granting. Credentials in project knowledge (B) is a data-exposure anti-pattern, not authentication. Disabling confirmations (C) removes the review step precisely where side effects are irreversible.
Q10: A. The documented practice is to validate a skill on representative real tasks and review its outputs before packaging it into a plugin, because a flawed skill inside a plugin scales the flaw across the whole team — exactly what happened. Restricting installers (B) changes who distributes the flaw, not whether it exists. Duplicated rules in global instructions (C) adds clutter without fixing the skill. Per-member plugins (D) defeats the purpose of encoding shared expertise.
Q11: C. The hallmark associate move when a need requires custom integration work — here, a connector that doesn't exist — is to scope the requirement and hand it off to developers, not attempt it in the chat surface. Hourly CSV exports (A) are the runner-up: feasible once, but unsustainable manual toil against hourly updates. Pasting rows (B) doesn't scale and invites stale answers. A connection string in instructions (D) does nothing — chat can't dial a database — and exposes credentials.
Q12: D. Project instructions are the persistent system prompt — the right home for a standing behavioral rule — while knowledge files hold the reference documents that rule points to. Stuffing a 25-page document into instructions (A) bloats the prompt and dilutes rule-following. All-files (B) leaves no persistent rule directing Claude to actually use the guide. (C) inverts the two roles.