Question 140
A project memory references a separate testing guide, but the team wants the guide edited independently. Which mechanism is useful?
- Reference the focused documentation from CLAUDE.md
- Replace test instructions with undocumented assumptions
- Encode the guide into an unrelated media asset
- Paste its complete contents into every new chat manually
Explanation
Correct Answer
A:A maintained linked file can preserve detailed test commands while leaving top-level guidance compact. Paths should remain valid in the repository.
Incorrect Answers
B:Generic inference may choose wrong commands or skip repository-specific checks. The explicit guide is valuable context.
C:Opaque artifacts are not a suitable instruction interface. The assistant needs readable, relevant documentation.
D:Manual repetition is error-prone and defeats reusable project configuration. The guide should remain an accessible file.
Question 141
A developer checks in an incorrect command in CLAUDE.md, and the whole team starts using it. What operational response best limits recurrence?
- Correct the shared file and add verification of the documented command
- Treat project instructions as immutable after a faulty command is detected
- Give the assistant more creative freedom to discover commands
- Hide the error by removing CI failure output
- Tell each developer to ignore the file indefinitely
Explanation
Correct Answer
A:Repository instructions propagate through workflows, so they need review like code. An executable check catches stale runbooks early.
Incorrect Answers
B:Versioned instructions should be corrected through review, just like code. An outdated guide remains a recurring source of failure.
C:Exploration may avoid one issue but introduces nondeterminism. The documented command should be fixed and tested.
D:Silencing test evidence conceals the defect. It does not repair inaccurate project guidance.
E:The shared source of truth would remain wrong. A central fix is more reliable than repeated local exceptions.
Question 142
A large repository needs shared coding standards plus a stricter instruction that applies only to files under the payments package. Which TWO configuration decisions fit? Choose TWO.
- Keep team-wide coding conventions in a versioned root CLAUDE.md
- Assume a CLAUDE.md rule can grant unapproved payment permissions
- Add a path-specific rule for the payments subtree
- Copy payment-specific rules into every unrelated package as global policy
- Store all shared rules only in one engineer's private conversation
Explanation
Correct Answers
A:The project file supplies reusable guidance to contributors and CI. Central maintenance reduces drifting personal copies of the same rules.
C:Scoped guidance can express specialized review requirements without applying them to unrelated packages. Hard security controls should still be enforced separately.
Incorrect Answers
B:Instruction files steer the assistant but do not authorize backend mutations. Actual permissions remain independently enforced.
D:Unnecessary duplication adds noise and can create conflicts. Path rules express the intended scope more precisely.
E:Ephemeral private context does not reliably reach collaborators or automation. The project needs versioned shared guidance.
Question 143
A team wants Claude Code to run a specific formatter after code changes. What should its instructions include to be actionable?
- Only the adjective beautifully formatted
- The exact command, the files or stage it applies to, and how to check success
- An instruction to skip all verification because formatting is cosmetic
- A long history of prior formatting disagreements with no current rule
Explanation
Correct Answer
B:Operational guidance works best when it is concrete and verifiable. A formatter rule should not be just a broad aspiration.
Incorrect Answers
A:Style language does not define a tool command or pass criterion. The model may guess an unsupported formatter.
C:Formatters can fail or modify files unexpectedly. The team should inspect relevant changes and exits.
D:Historical debate increases noise. The active standard should be unambiguous.
Question 144
A developer notices that personal CLAUDE.md instructions conflict with mandatory repository security rules. What is the best practice?
- Assume the most recently edited file automatically grants broader privileges
- Delete the repository's policy to accommodate one developer's preference
- Apply project policy and enforce security outside instructions
- Ask Claude to choose whichever rule sounds easier
Explanation
Correct Answer
C:Instruction files help guide behavior, but mandatory controls belong in permissions and hooks. Personal convenience cannot override security requirements.
Incorrect Answers
A:File modification time is not an authorization model. Tool access and organizational policy still govern actions.
B:Removing shared protections undermines team standards. Changes should go through approved review.
D:Convenience is not an accepted security criterion. Conflicts require explicit governance.
Question 145
A team includes many external links inside CLAUDE.md and wants offline reliability. What change would help?
- Make every required workflow depend on an untrusted public pastebin
- Keep essential instructions and critical commands in versioned local files
- Assume all linked websites will remain reachable from every CI runner
- Delete all local project commands and rely entirely on model memory
Explanation
Correct Answer
B:Offline work cannot depend on pages that may be unreachable. Local authoritative guidance supports predictable behavior and reproducible review.
Incorrect Answers
A:External mutable pages weaken provenance and availability. Versioned project files are safer for core instructions.
C:Network availability varies, and links can change. Critical workflow details should not require unverified access.
D:Model recall may omit environment-specific instructions. The local source remains valuable.
Question 146
A development team repeatedly requests the same security review checklist with identical acceptance criteria. Which Claude Code feature packages that reusable workflow?
- Disable all supporting files to keep prompts short
- Create a project skill with a documented invocation path
- Add the entire checklist to each individual chat by hand
- Rename the repository so Claude assumes a review process
Explanation
Correct Answer
B:A skill can preserve procedure and relevant resources for repeated use. Team ownership allows versioning of the review criteria.
Incorrect Answers
A:The checklist may need concrete rules and examples. Removing its supporting context undermines repeatability.
C:Repetition is error-prone and hard to update consistently. A reusable workflow artifact better meets the request.
D:Repository names do not define executable guidance. The workflow needs explicit instructions.
Question 147
A teammate creates .claude/skills/release-check/SKILL.md with a documented workflow. What is its main purpose?
- Provide reusable, discoverable instructions for that named workflow
- Automatically replace every repository test with a model opinion
- Serve as a compiled binary used by the application
- Grant permanent administrator credentials to Claude Code
Explanation
Correct Answer
A:A skill groups task-specific behavior for invocation when relevant. It is not a replacement for file permissions or application authorization.
Incorrect Answers
B:A workflow can direct tests, but it cannot make actual test execution unnecessary. Verification remains essential.
C:Markdown instructions are read by the assistant, not executed as the project's program.
D:A skill is an instruction artifact, not a credential store or privilege grant. Execution rights remain separately controlled.
Question 148
A workflow is invoked with a ticket identifier that changes each time. How should the reusable skill accept this context?
- Specify an input argument or placeholder for the ticket id and validate it
- Hard-code a real customer's ticket into the shared skill
- Ask the model to invent an identifier when none is provided
- Embed a private access token beside the reusable ticket argument
- Make the skill work only when the ticket number matches an example
Explanation
Correct Answer
A:Parameterized instructions let one workflow operate on different cases without duplicating files. Invalid or missing identifiers should be handled explicitly.
Incorrect Answers
B:Embedding case-specific data creates accidental reuse and privacy concerns. Inputs should be supplied at invocation.
C:Fabricated IDs could target wrong records. The workflow should request or resolve a valid value.
D:A workflow parameter should identify the task, not carry secrets in shared instructions. Use secure runtime authorization.
E:A fixed example is not a reusable operational interface. Parameterization is the desired property.
Question 149
A team wants its release-review skill to inspect files but never run deployment commands. What configuration assists least privilege?
- Ask a reviewer to undo deployments after the fact
- Give the skill unrestricted shell permissions to simplify invocation
- Declare allowed tools and enforce deployment permissions
- Trust the skill's title as proof that it cannot deploy
Explanation
Correct Answer
C:Skill tool metadata can narrow the intended capabilities. The executor must still prevent forbidden actions independent of instructions.
Incorrect Answers
A:Recovery is weaker than prevention and may not reverse external effects. Restrict mutation tools up front.
B:Broad execution access is unnecessary for review and risks an accidental deployment. The capability surface should match the task.
D:A descriptive name does not constrain actual tool access. Explicit permission control is needed.
Question 150
A project skill grows too large because it embeds all examples, policy manuals and implementation notes in one prompt. What structure is better?
- Convert all instructions into a hidden tool result
- Use a focused SKILL.md with referenced support files
- Remove the workflow steps and keep only a title
- Clone the whole document into every skill in the project
Explanation
Correct Answer
B:Separating core procedure from detailed resources reduces irrelevant context and eases updates. The skill should expose a clear entry point.
Incorrect Answers
A:Tool results are observations, not a durable skill definition. The procedure belongs in a maintained instruction file.
C:Without concrete instructions, the skill cannot consistently guide execution. Shorter should not mean empty.
D:Duplication increases maintenance and conflicting versions. Shared supporting resources are preferable.