Skip to main content

Extending and Contributing to NanoClaw

1. Start with the skill model first​

NanoClaw's contribution guidance is refreshingly explicit:

do not add broad new features to trunk if they belong as skills

That means your first design decision is whether the change belongs in:

GoalFirst place to look
host runtimecore src/ host files
provider or channel moduleskill-installed branch or module
per-agent behaviorgroup-level workspace or container layer
docs and migrationdocs and migration scripts

2. Why this matters​

The project wants trunk to remain:

  • small,
  • understandable,
  • and safe to fork.

If you ignore that goal, you will probably fight the repo's philosophy.

3. Best first contributions​

  1. docs and architecture clarification,
  2. migration and install polish,
  3. skill examples,
  4. targeted host-runtime fixes,
  5. isolation-model improvements.

4. Think in maintainable forks​

A good NanoClaw contribution should help users keep their fork lean and comprehensible, not make the main branch harder to audit.

5. Before opening a PR​

Ask whether the capability belongs in trunk, a skill branch, or the user's own fork. In NanoClaw, that choice is part of the architecture.