Concepts

Explore the models and boundaries behind skills, composition, capabilities, harnesses, and governed knowledge.

Start here when you want to understand why Adaptive Skills is structured the way it is. These concepts explain the system; they do not add a second runtime or change how a skill executes.

Skills

Skills are portable micro-execution assets. Each skill carries a focused method, activation triggers, optional modules, verification expectations, failure signals, and handoff guidance.

Start with:

• Skill model — anatomy and selection discipline.
• Skill categories — current organizational taxonomy.
• Domain taxonomy — portable skills versus domain packs.
• Lean skill doctrine — why skills stay small and composable.

Composition

Composition describes how multiple skills participate in one task without becoming a hidden workflow engine.

Start with:

• Execution patterns — supported participation patterns.
• Looping models — bounded iteration and stop conditions.
• Workflow recipes — practitioner-oriented examples.
• Skills in orchestrated workflows — AletheIA stage participation.

Capabilities

Capability metadata helps consumers discover and route to skills. It is an advisory overlay; the canonical executable assets remain under skills/.

Start with:

• Capability model — definitions and layer boundary.
• Capability graph — relationships among capabilities.
• Capability routing boundary — what routing may and may not decide.
• Execution modes — declared ways capabilities participate.

Harnesses

Harnesses make skills available in a runtime. They own local loading, permissions, tool access, and execution mechanics; skills retain their portable method and declared requirements.

Start with:

• Skill and harness boundaries.
• Using skills inside harnesses.
• Harness requirements.
• Operational runtime.

Knowledge

Knowledge-aware skills can use governed source material without embedding proprietary content or silently treating untrusted text as instruction.

Start with:

• Skill knowledge boundaries.
• Declaring knowledge dependencies.
• Using proprietary frameworks safely.

How the families connect

1. A skill defines a portable execution method.
2. Composition defines how that method can participate with other bounded methods.
3. Capability metadata improves discovery without replacing the skill.
4. A harness supplies local runtime mechanics and authority boundaries.
5. Governed knowledge supplies source-backed context when the skill declares it.

Next steps

• Use the ecosystem map for a compact system view.
• Read Adaptive Skills and AletheIA for the macro/micro boundary.
• Return to Choose the right skill when your goal is practical selection rather than architecture.