Give coding agents safe context for Unreal projects
ForgeLore gives assistants the same product surface humans use: repository status, recent activity, locks, summaries, connection guidance, dry-runs, and scoped write actions. Keep access explicit and make every agent respect Unreal asset locks.
Start assistants with the scopes they need for the task. High-impact actions still require a human confirmation step.
Codex
Use AGENTS.md so every Codex task starts with project rules, test commands, and Unreal asset guardrails.
Claude Code
Use CLAUDE.md memory to keep project conventions, lock rules, and verification expectations close to the code.
Cursor
Use project rules so agent mode understands what to avoid and when to ask a human first.
MCP clients
Use ForgeLore tools for repository status, locks, activity, and connection info within explicit scopes.
Copy this starter prompt
Paste it into Codex, Claude Code, Cursor, or another assistant before assigning work.
Agent checklist
Use this before granting an assistant access to a project.
Codex AGENTS.md
Claude Code CLAUDE.md
Cursor project rule
Generic MCP instruction
Use this when an assistant connects through ForgeLore MCP tools. Tool access should be scoped, revocable, and guided by the capability manifest.