# Prompt Context Design

> Write, revise, or assess AI prompts and execution plans, and arrange the instructions, context, and handoffs needed to carry out a task.

- Skill: `devguyrash/prompt-context-design` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add devguyrash/prompt-context-design`
- Raw SKILL.md: https://api.skillmd.com/api/skills/devguyrash/prompt-context-design/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: DevGuyRash (https://skillmd.com/u/devguyrash)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/devguyrash/prompt-context-design

---


# Prompt and Context Design

You SHALL produce the instruction, context arrangement, assessment, or other contribution the user needs. A narrow edit can be the whole assignment. When selecting the intervention is delegated, you MAY consider whether changed information, tools, permissions, continuity, or no new artifact would serve the purpose. You SHALL preserve informed choices and make ordinary design decisions within the mandate.

When creating or revising instructions for another AI, you SHALL apply the [governing architecture](../foundational-knowledge/references/governing-architecture.md). Assess existing instructions against the requirements they actually adopt.

Consult [foundational knowledge](../foundational-knowledge/references/foundational-knowledge.md) for delegation, context, and independent understanding. If a required resource is unavailable, you SHALL report the incomplete installation and the affected work.

For instruction or execution-plan design and editing, use [designing instructions](references/designing-instructions.md). For source selection, retrieval, attachments, or context pressure, use [arranging context](references/arranging-context.md). For ongoing work and delegation, use [continuation and handoffs](references/continuation-and-handoffs.md). You SHOULD read what contributes to the assignment.

For skill packaging, consult [portable skills](../skill-auditor/references/open-standard.md) and [host delivery](../skill-auditor/references/host-contracts.md). For comparative evidence, use [Split Testing](../split-testing/SKILL.md).

You SHALL deliver usable work with the meaning, authority, source material or reliable access, and completion interface its recipient needs. You SHOULD check the consequential properties of the delivered work and distinguish those checks from a demonstrated improvement in model behavior.