You are a refining a prompt over several iterations. The output drifts further from the requested template with each change.
Which step is most appropriate?

• A. Stop iterating and accept the most off-target version.
• B. Continue making large simultaneous changes until the output happens to align.
• C. Discard all prior versions and start with no record of what worked.
• D. Revert to the version of the prompt that produced the closest match and adjust from that baseline.

You are working as a process consultant integrating a pre-built Skill into a recurring team workflow.
Which two practices best support reliable integration? (Select two.)

• A. Apply the Skill outside its documented scope so the team can stretch its use across several unrelated workflow steps.
• B. Remove all human review from steps that invoke the Skill, relying on the Skill's packaged output as final.
• C. Document which workflow steps invoke the Skill and what human review remains, so accountability and quality gates stay visible to the team across recurring runs.
• D. Define the inputs the Skill expects and the outputs the team will use, so each workflow step has clear hand-off points around the Skill.
• E. Bypass the published Skill version by using an unreleased build pulled directly from outside the approved distribution channel.

You are an operations lead evaluating prompting strategies that teammates have proposed for various task types.
Which two strategies represent appropriate matches between strategy and task type? (Select two.) Each correct answer presents part of the solution.

• A. rigid templated prompting for casual customer-feedback summarization with no schema
• B. rigid templated prompting for early-stage creative naming and tagline ideation
• C. open exploratory prompting for legal contract redlining against a clause library
• D. rigid templated prompting for monthly compliance reports against a fixed schema
• E. open exploratory prompting for early-stage strategic brainstorming on a new market

A knowledge worker is preparing to use the Gmail Connector with Claude for the first time.
Which statement most accurately describes the Connector's behavior at the user level?

• A. Claude searches and reads emails the user has access to and labels them according to user-defined rules, with reply drafting handled in a separate workflow.
• B. Claude searches and reads emails the user has access to and sends replies on the user's behalf for low- risk messages, with higher-risk messages routed to the user for approval.
• C. Claude searches and reads emails the user has access to and drafts replies for the user to send manually, with each action requiring explicit user approval.
• D. Claude searches and reads emails the user has access to in a read-only mode, and the user must compose any reply manually outside the Connector.

You are a knowledge manager structuring source material in the context window for a long Claude evaluation task.
Which two practices produce a more effective use of the context window? (Select two.) Each correct answer presents a complete solution.

• A. Place the most reference-critical material near the beginning or end of context.
• B. Paste every available document into context regardless of relevance to the task.
• C. Mix sources together in a single block so the model can synthesize them freely.
• D. Mark each source clearly with delimiters so the model can attribute claims correctly.
• E. Repeat the same instruction at the start, middle, and end to reinforce it.