Skip to main content

Creating Prompts

Folder: .github/skills/tsh-creating-prompts/ Used by: Copilot Engineer

Provides templates, guidelines, and a structured process for building custom prompt files (.prompt.md) that trigger specific workflows routed to the right custom agent and AI model.

Core Design Principles​

  • Separation of concerns — A prompt file defines WHAT workflow to execute. It must NOT define WHO the agent is (agent territory) or embed reusable domain knowledge (skill territory).
  • Workflow focus — A prompt routes work to a specific agent, targets a specific AI model, and describes workflow steps for a specific task.
  • Minimal duplication — Prompts reference skills and agents; they don't duplicate their content.

Prompt File Structure​

SectionPurpose
YAML Frontmatteragent, model, description fields
Workflow StepsOrdered steps the agent should follow
Input/OutputWhat the prompt expects and what it produces
Required SkillsSkills the agent must load for this workflow

Key Guidelines​

  • Every prompt must specify an agent field routing to a tsh- prefixed agent name
  • Every prompt must specify a model field (e.g., Claude Opus 4.6)
  • Prompt filenames follow tsh-<action>.prompt.md convention
  • Prompts should not redefine, override, or contradict the agent's identity

Validation Checklist​

  • YAML frontmatter includes agent, model, description
  • Agent reference uses tsh- prefix
  • No personality or behavioral content (agent territory)
  • No reusable workflow steps (skill territory)
  • No coding standards (instructions territory)

Connected Skills​

  • tsh-creating-agents — For creating the agent a prompt routes to.
  • tsh-creating-skills — For creating skills referenced by the prompt.