From Messy Agent Configs to One Source of Truth

Where this fits

You're a developer who's been experimenting with AI coding agents. Over months, your project accumulated a .claude/ folder, a .codex/ folder, a AGENTS.md that half-works, and a custom skills/ directory with inconsistent naming. Claude Code follows one set of rules, Codex ignores others, and Grok doesn't even know where to start. Every time you switch tools, the agent re-learns your project from scratch—or worse, contradicts itself. This is the exact problem dbs-agent-migration solves: it turns that fragmented, semi-migrated state into a single, maintainable workspace that all three tools can share.

Why agents benefit

• Audits rule files across hosts – automatically identifies conflicting instructions in .claude/, .codex/, AGENTS.md, and Grok configs, so agents don't get ambiguous directives.
• Establishes a true source of truth – picks one canonical skill set and generation point, then makes every other host reference that source instead of duplicating logic.
• Normalizes naming conventions – renames skills and commands to a consistent pattern, so triggers like "migrate to Codex" work regardless of which tool you're currently in.
• Generates bridge files – creates glue rules that translate between host-specific syntaxes (e.g., Claude's .claude/commands and Codex's AGENTS.md), letting agents read each other's context.
• Validates the final structure – runs a verification pass to catch missing references, orphaned files, or stale rules before you commit.

Practical scenarios

The legacy project that switched agents twice

You migrated from Claude Code to Codex last quarter, but the old .claude files are still there, half-updated. Codex occasionally reads them by accident. Running this skill audits both rule sets, merges the valid ones, and generates a bridge so Codex ignores stale Claude-specific directives while still respecting shared project conventions.

A greenfield repo with three collaborators using different tools

Your team has one dev on Claude Code, one on Codex, and one experimenting with Grok. Each has written their own AGENTS.md and skills, causing merge conflicts and duplicated effort. The skill unifies all three into a single source-of-truth skill directory, then generates bridge files so each tool loads the same rules in its native format.

The “my agent workspace is a mess” emergency

You're launching a new feature and the agent keeps misinterpreting commands because skill names clash across hosts. A quick /dbs-agent-migration triggers an audit, renames conflicting skills with a clear prefix, and regenerates the bridge files. Within minutes, every agent sees the same normalized structure.

Add it to your agent workflow

Install it directly from your agent's command line:

npx skills add Pluviobyte/rnskill --skill dbs-agent-migration

Then trigger the migration with a natural-language prompt like: "统一 AGENTS.md" or simply "migrate to Claude Code". For a full migration, just run:

/dbs-agent-migration

The skill will walk through audit → identify source → normalize → generate bridges → validate, and leave you with a clean, coherent workspace.

Compare before adopting

Before committing, compare this skill against a manual cleanup script or a simple copy-paste of AGENTS.md. Look for:

• Quality signals: Does it actually detect conflicts, or just concatenate files? Test on a repo with known contradictions.
• Maintenance freshness: Check the GitHub repo's recent commits and issue resolution speed—stagnant projects fall behind new agent versions.
• Alternatives: Some tools only handle one host; this one explicitly targets three. If you're only on Claude Code, a simpler skill might suffice.
• Workflow fit: Does your team use all three tools? If not, the bridge generation overhead may be unnecessary.

Why it is worth tracking

With 1,381 stars and active development, this skill addresses a real pain point that's only growing as multi-agent workflows become standard. It's worth tracking because it's not just a one-off migration script—it's a maintainable framework for keeping your agent workspace consistent long-term. Evaluate it when your project starts showing signs of config sprawl, or before your team expands to a second agent tool. For a deeper dive, check the skill's official listing on OpenAgentSkill.