Feature Idea: Multiple Rule Source Directories (--ruler-dirs)

[modanets]

First — huge thanks for building ruler. The idea of a single source of truth that distributes to 30+ agents is exactly the kind of DX tooling this ecosystem needs.

The use case: third-party skills as packages

As the AI tooling ecosystem matures, teams will want to share and reuse skills across projects — either their own internal libraries or community-published packages. The natural distribution mechanism for this is npm, composer, or git submodules.

Right now, if I want to pull in a third-party skills package, I hit a wall: ruler only looks at .ruler/ in the project root. That directory name is also hardcoded, which means:

• npm packages land in node_modules/my-skills/ — ruler can't see them
• composer packages land in vendor/my-skills/ — same problem
• Even with git submodules, I'm locked into naming the submodule .ruler/ (or a subdirectory of it)

The ask: --ruler-dirs (plural) as an array

The ask is simple: implement a --ruler-dirs configuration property that accepts multiple source directories:

# Mix your local rules with one or more third-party packages
ruler apply \
  --ruler-dirs .ruler \
  --ruler-dirs node_modules/@myorg/ai-skills \
  --ruler-dirs node_modules/@community/typescript-skills

Or via ruler.toml:

ruler_dirs = [
  ".ruler", // this one is there by default for backward compatibility
  "node_modules/@myorg/ai-skills",
  "node_modules/@community/typescript-skills"
]

Ruler would merge/concatenate all sources in order, then distribute to agents as it does today. The output experience is identical — only the inputs change.

Why this is a natural fit for the ruler

The killer insight here is that ruler already does the hard part — merging multiple .md files and distributing to 30+ agents. Supporting multiple input directories is a small lift that unlocks a whole ecosystem of reusable AI rules.

Would love to hear if this fits the roadmap!