AI-agent standardization¶

Repo Standards uses Rulesync to standardize AI-agent and editor instructions across repositories.

Policy¶

AGENTS.md is the universal generated AI-agent instruction file.

Generated AI/editor files should not be hand-edited as source material. Change the source files and regenerate outputs instead.

Canonical source in downstream repositories:

.repo-policy.yml
rulesync.jsonc
.rulesync/rules/*.md

Generated outputs:

AGENTS.md
.cursor/rules/*.mdc
.agents/rules/*.md
.agents/memories/*.md

During migration, the wizard detects:

AGENTS.md
CLAUDE.md
.cursorrules
.cursor/
.agents/
.antigravity/

These files and directories may contain old or generated AI/editor instructions. The wizard offers to replace them with fresh Rulesync output.

Because this is destructive, the wizard must require explicit confirmation before deleting or replacing them.

Recommended confirmation text:

replace-ai-files

CLAUDE.md is treated as legacy/non-standard unless Repo Standards intentionally adds and documents a supported Rulesync target for it.

Default behavior:

Repo Standards maintains canonical rule templates in:

ai/rules/*.md

Downstream repositories copy selected rules to:

.rulesync/rules/*.md

Language, framework, deployment, and governance modules may add their own rules.

Example rule layers:

00-org.md
10-typescript.md
20-python.md
30-cloudflare.md
31-firebase.md
40-home-assistant.md

After running Rulesync, review generated outputs before committing:

find AGENTS.md .cursor .agents .rulesync -maxdepth 4 -type f -print | sort

Expected generated outputs depend on enabled Rulesync targets.

A standards migration PR should make clear: