How we write code
- Always
- TypeScript strict. Small functions. No new dependencies without asking.
- Never
- Commit secrets or push to main.
Use cases / Agent rules
One set of instructions every agent reads, on every machine, in every tool.
Read flie.sh/use-cases/agent-rules.md and move my agent rules into it.Paste into your agent, in a repository that has rules. See what it sets up
An example: your agent writes “How we write code” in the folder Agent rules.
The problem
.cursor/ folders in one study’s corpus of coding-agent configuration.
MSR 2026 study of coding-agent configurationCopilot instruction files in the same corpus, and 16,600 .cursorrules: a file per tool for the same rules.
MSR 2026 study of coding-agent configurationDevelopers build tools just to carry agent rules and memory from one machine to the next.
agent-sync, GitHubWhat your agent sets up
# flie.sh/use-cases/agent-rules.md Make a folder "Agent rules" with how we write code.md how we test.md One file per topic; each repository points here.
Plain Markdown on flie’s site. Read the whole recipe.
Agent rules/ README.md the rules how we write code.md one file per topic how we test.md
Every agent reads the README first. Edit it and you change the rules for all of them.
# AGENTS.md or CLAUDE.md + Read and follow the rules in the flie folder "Agent rules" (fld_…) before you start; its README says how.
So every future session keeps it current, not just this one.
Who reads it
Same rules.
Same rules.
Same rules.
Reads them in the browser.
Read flie.sh/use-cases/agent-rules.md and move my agent rules into it.Free · Every version kept · Install it yourself