Skip to content

Custom Prompting

Improve AI upgrade guidance by providing concise repository context and clear PR information. The service automatically gathers these inputs when generating tickets and implementing changes.

Repository context

Add a single, top‑level guidance file to your repo (first match wins):

  • agents.md (preferred)
  • AGENTS.md
  • docs/agents.md
  • .codelogic/agents.md
  • .codelogic/context.md
  • .ai/context.md
  • .ai/agents.md
  • CODELOGIC_AI.md

What to include (keep it short):

  • Build & test commands
  • Project layout and module mapping
  • Required tools/runtimes
  • Policies (commit conventions, test expectations)

Example outline:

# Repository AI Context

## Build & Test
- Build: ./scripts/build.sh
- Test: ./scripts/test.sh

## Project layout
- Backend: ./service
- Web UI: ./ui

## Policies
- Do not commit secrets
- Keep changes atomic; include tests when changing logic

Pull request context

The service reads PR title/body and recent comments (issue and review). Use the PR body to:

  • Link release notes (e.g., Renovate/Dependabot)
  • Describe known migration steps or constraints
  • Call out non‑standard build commands or modules

How context is used

When implementing tickets, CodeLogic builds a single prompt by prepending:

  1. Pull request metadata (title/body/comments)
  2. Repository context (from your guidance file)
  3. The generated ticket content

This helps the AI follow your build/test conventions and repository structure.