Anatomy of a .kcp Package

Schema

Anatomy of a KCP Package

Every .kcp package is a structured collection of cognitive blocks. Each block is independently useful and can be consumed by retrieval systems, agents, knowledge graphs or fine-tuning pipelines.

Metadata

Identity, version, source, language and quality scores for the package.

Core Intent

The primary purpose, intended user, intended agent use and transformation goal of the source.

Domain Map

Main domain, subdomains, adjacent domains and explicitly excluded domains.

Entity Graph

People, roles, tools, concepts, metrics, states and objects extracted from the source.

Concept Graph

Key ideas as connected semantic nodes with dependencies, contradictions and supporting relationships.

Principles

Foundational statements with conditions of application and rationale.

Heuristics

Tacit expertise turned into practical rules of thumb for agent reasoning.

Decision Rules

Conditional decisions with required context, output actions and failure modes.

Procedures

Step-by-step processes with inputs, outputs, success and failure criteria.

Patterns

Recurring situations, signals, mechanisms and response strategies.

Anti-patterns

Failure modes with warning signals and replacement behavior.

Causal Chains

Cause, mechanism, effect, secondary effects and intervention points.

Contextual Triggers

When-then activations: signals that should make agents recall specific knowledge.

IF-THEN Rules

Compact conditional logic with rationale and confidence.

Exceptions

Edge cases that override general rules with modified actions and explanations.

Mental Models

Reusable frames of reasoning, with usage and non-usage conditions.

Operational Playbooks

End-to-end agent runbooks: objective, steps, tone, tools and failure modes.

Agent Q&A

Reference question/answer pairs for evaluation, tutoring and onboarding.

Retrieval Chunks

Standalone, dense, context-rich units optimized for semantic search.

Atomic Units

One-idea statements suitable for embeddings, evaluation and memory systems.

Agent Instructions

Behavior, reasoning, response, forbidden actions and tool guidance.

Knowledge Limits

Missing context, weakly supported claims, biases and items needing human review.

Source Traceability

Links every generated item back to its original source location and excerpt.

Comparison

The same knowledge, two formats

A human writes a paragraph. A .kcp package turns the same content into typed cognitive blocks an agent can query.

Good for humans

To approve a refund above $500, the support agent should first check the customer's order history. If the customer is loyal and the issue is clearly our fault, escalate to a manager — but never before confirming the product was actually shipped. In edge cases involving fraud signals, hold the refund and open a review.

.kcp · Structured for agents

entity

rule

procedure

heuristic

exception

atomic_unit

entity: support_agent, customer, refund, manager
rule:
  if: refund.amount > 500 AND fault = internal
  then: escalate(manager)
  requires: order.shipped = true
procedure:
  - check(customer.order_history)
  - verify(product.shipped)
  - escalate_if(fault = internal)
heuristic: loyal customers tolerate slow refunds, not silence
exception:
  when: fraud_signal detected
  action: hold_refund AND open_review
atomic_unit: refunds above 500 require manager approval