Progressive Disclosure

This repository uses progressive disclosure so agents load the minimum standing context, then pull depth on demand via skills, scoped rules, and generated docs.

Layers

Layer	Owner	Loads when
Global instructions	instructions/global.md	Always (cross-harness)
Skill descriptions	skills/*/SKILL.md frontmatter	Always (descriptions only)
Scoped rules	.claude/rules/, .cursor/rules/	Path match
Skill bodies	skills/*/SKILL.md body	Skill invoked
Generated docs	wagents docs generate	Human browse / MCP docs-index

Token efficacy

See Harness config — token efficacy for one-tool-per-layer policy.

Skill topology

Source topologyGenerated from the same custom, installed, and curated external skill index used by the catalog.