msia-prompts

Preventing duplicate rules across agent prompts requires a canonical location map that assigns every rule type a single home. This governance approach eliminates conflicts by enforcing structure and preventing redundant instructions across your prompt files.

A canonical location map in prompt engineering is a governance structure that gives every rule a single, true home within your repository. It prevents duplicates and conflicts across core modules and mode files by strictly managing where specific instructions must reside.

Managing token budgets when editing agent prompts requires dedicated governance workflows that track and allocate token limits across core modules and mode files. A pre-edit checklist ensures changes remain within budget while maintaining structural safety and preventing anti-patterns.

Safe edits on agent prompt markdown files require loading a governance skill before modifying any file under agent/prompts/**/*.md. This enforces a pre-edit checklist and an auditable workflow, ensuring changes do not introduce anti-patterns or duplicate rules.

The best way to enforce safety checks and structure in prompt engineering is through an auditable prompt-systems workflow. This workflow applies lint-tests and a pre-edit checklist to guide changes, ensuring all edits follow canonical location rules and avoid known anti-patterns.

You need canonical prompt governance when your agent system grows large enough that rule duplication and conflicts across markdown files become unmanageable. Implementing a single canonical home per rule type with token budgeting prevents these structural issues before they degrade agent performance.