Skip to content
Adaptive Skills Docs
Esc
↑↓navigate↵open⌘Jpreview
On this page

AI Discovery & Agent Experience

An experimental five-skill domain pack for evidence-backed discovery, representation, measurement, and safe agent actions.

This experimental domain pack helps teams make public knowledge easier to interpret, search foundations easier to diagnose, generative behavior measurable, and agent actions safer to design. It provides five domain skills without adding a crawler, analytics product, agent runtime, or integration.

Start with the decision

  1. Name the property, user task, decision owner, and evidence you can access.
  2. Use Choose a skill to select the smallest sufficient contract.
  3. Follow a workflow only when several evidence boundaries interact.
  4. Use the evidence and measurement contract for every material finding.
  5. If access, authority, or proof is missing, record the gap and hand the decision to its owner.

The five skills

Skill Use it for
knowledge-entity-representation Distinguish people, organizations, brands, methods, products, claims, and relationships.
search-indexability-optimization Diagnose access, crawl, render, indexability, canonicalization, architecture, and structured representation.
ai-discovery-measurement Design repeatable query, retrieval, citation, representation, or task observations.
generative-visibility-optimization Compare repeated answer behavior with canonical evidence and form bounded improvement hypotheses.
agent-capability-actionability Define tasks, preconditions, authorization, side effects, confirmation, and recovery.

What the pack does not promise

The pack never guarantees indexing, rankings, citations, traffic, conversion, adoption, or successful agent execution. Technical formats such as schema, llms.txt, WebMCP, MCP, A2A, and OpenAPI remain conditional choices. Production changes and high-impact actions require their normal owners and controls.

Continue

Was this page helpful?