Prompt and Context Engineering

CCDV-F · 22 questions

• A permitting agent receives verbose GIS tool dumps that threaten the next reasoning turn. What context step fits?
• A 311 session nears the context-window limit during a long resident chat. What management step fits?
• A court research workflow must keep search noise from drifting into drafting context. Which isolation pattern fits?
• A transit trip planner refreshes arrival predictions but old values remain in context. What hygiene step fits?
• A housing case bot must keep working memory lean across eligibility turns. Which selection practice fits?
• A public-health assistant faces multi-page lab PDFs before Claude deliberates. Which compaction strategy fits?
• A parks multi-step agent moves from inspect to notify and carries noisy tool residue. What context practice fits?
• A school-district agent must handle student PII and curriculum-policy text carefully. Which context design fits?
• A city needs Claude to cite ordinance sections instead of vague paraphrases. Which prompt engineering step fits?
• A 311 team must keep immutable city policy separate from free-text resident messages in the prompt. Which placement practice fits?
• A transit team needs rider-alert messages to keep a stable tone and length. Which prompt engineering technique fits?
• A court clerk task should list only missing filing fields, nothing else. Which prompt constraint fits?
• A housing eligibility prompt confuses income thresholds on eval. What refinement practice fits?
• A water utility builds prompts from free-text resident messages. Which assembly step fits?
• A library wants catalog-note style rules to stick when Claude writes output. Which placement approach fits?
• A school IEP helper must avoid unauthorized medical diagnosis wording. Which prompt rule fits?
• Emergency PIO staff say Claude boilerplate is too long. Which prompt practice fits?
• A permitting API receives a Claude JSON checklist destined for the case file. Which output handling step fits?
• A 311 parser gets confident but malformed ticket JSON from Claude. What defensive handling fits?
• A transit integrator needs ETA fields that downstream systems can parse. Which output approach fits?
• A housing UI must not show Claude dollar amounts that fail numeric checks. Which gate fits?
• A court export job consumes Claude summaries for legal clerks. Which defensive parse step fits?