Building blocks and capability families

Understand direct skill use, optional family entrypoints, and selective composition.

Adaptive Skills is a library of small, inspectable building blocks. A skill can be used directly when the task is clear, or through a family entrypoint when the request is broad, ambiguous, or composed.

Two equivalent entry paths

Direct building block

Use a known skill when its contract matches the task:

"Review this page's indexability"
→ search-indexability-optimization
→ bounded findings and verification

Family entrypoint

Use a family when the user knows the outcome but not the internal capability:

"Improve our SEO"
→ AI Discovery & Agent Experience
→ diagnose the dominant uncertainty
→ select one primary building block
→ add supporting blocks only when evidence requires them

The family is a guided composition, not a replacement for its members.

Selective composition

Families and skills must not load every descendant by default. New evidence may justify a module or another skill during the work. Each activation should state its reason, expected input, output, and verification. A rejected building block should be named when the choice could otherwise be ambiguous.

Boundaries

• A skill owns a bounded method and reviewable output.
• A module adds conditional depth inside a skill.
• A family or macro-capability describes a useful composition of building blocks.
• A harness owns tools, loading, authentication, execution, and logs.
• AletheIA owns macro intent, authorization, gates, and consequential decisions.

The family does not become an autonomous runtime, approve decisions, or change external state. Advanced users may always bypass the family and invoke a canonical skill directly.

When to choose each path

See the skill catalog for direct discovery and the capability model for the underlying composition model.