Structured Data
Capability Metadata
| Field | Value |
|---|---|
| ID | CAP-104 |
| Version | v1.27 |
| Domain | Specialist Domains |
| Group | Knowledge Packs |
| Status | Documented |
| Related | CAP-098, CAP-099, CAP-103 |
| Format | YAML frontmatter + structured sections |
| Storage | .claude/knowledge-packs/ |
Configuration Reference
YAML frontmatter fields: name, domain, version, observation_hooks, event_subscriptions, promotion_threshold
Section structure: overview, concepts, exercises, skills
name: apollo-agc-basics
domain: apollo-agc
version: "1.0"
description: >
Foundation knowledge for Apollo Guidance Computer programming.
Use when: working with AGC assembly, DSKY interaction, or navigation.
observation_hooks:
- pattern: "agc|dsky|verb noun|tc |ccs "
confidence: 0.8
event_subscriptions:
- event: "agc.instruction.executed"
handler: record_pattern
promotion_threshold:
observations: 3
confidence: 0.75
sections:
- overview
- concepts
- exercises
- skills
Technical Specifications
Format basis: GSD Skill format extended with domain-specific metadata
Observation hooks: regex/keyword patterns with confidence threshold — trigger knowledge pack activation when domain-relevant content appears (see CAP-098)
Event subscriptions: route knowledge-pack events to skill pipelines without modifying core pipeline code (see CAP-099)
Promotion threshold: minimum observations and confidence required before pack generates a promoted GSD skill (see CAP-103)
Storage path: .claude/knowledge-packs/{domain}/{name}.yaml
Overview
The Knowledge Pack Format is the container specification that all other knowledge pack capabilities build upon. It extends the GSD Skill Creator's skill format to handle structured domain expertise — not just workflow patterns but conceptual knowledge with exercises, examples, and cross-domain connections.
A knowledge pack is a YAML document with two major parts: a frontmatter section that declares how the pack interacts with the observation and event systems, and a content section that organises domain knowledge into overview, concepts, exercises, and promoted skills. The frontmatter is machine-readable metadata; the content sections are human-readable documentation that adapts to the learner's mastery level.
Understanding the Knowledge Pack Format is prerequisite to understanding every other capability in the knowledge pack subsystem. CAP-098 through CAP-103 all operate on knowledge packs defined in this format — the observation hooks declared in the frontmatter (CAP-098), the event subscriptions routed by the bridge (CAP-099), and the promotion threshold used when generating permanent skills (CAP-103) are all fields specified here.
History
Knowledge packs emerged from observing that specialist domains — Apollo AGC, Amiga Heritage, electronics, Minecraft — required more structure than individual GSD skills. A skill captures a workflow pattern; a knowledge pack captures the conceptual background needed to understand and apply that workflow in a specialist domain.
The format was designed in v1.26 as a deliberate extension of the existing skill YAML format, adding three domain-specific capabilities: observation hooks (so the pack knows when to activate), event subscriptions (so domain observations route to the right handlers), and promotion thresholds (so demonstrated mastery generates permanent skills automatically). The backward-compatible design means any existing skill file can be upgraded to a knowledge pack by adding the new frontmatter fields.
Key Concepts
Think of a knowledge pack like a textbook with smart bookmarks. The textbook itself contains everything you need to learn about a subject — overview, key concepts, exercises, and skills. But the smart bookmarks (observation hooks) watch what you're working on and automatically open the book to the right page when you encounter relevant content.
The format is the specification for what those smart textbooks look like: how to write the table of contents (frontmatter), how to structure the chapters (sections), and how to set the threshold for when you've learned enough to move on (promotion_threshold).
Usage Examples
Example 1: Minimal knowledge pack for a specialist domain
# .claude/knowledge-packs/minecraft/redstone-basics.yaml
name: redstone-basics
domain: minecraft
version: "1.0"
description: >
Foundation knowledge for Minecraft Redstone engineering.
Use when: designing circuits, troubleshooting signal delays.
observation_hooks:
- pattern: "redstone|repeater|comparator|piston"
confidence: 0.75
promotion_threshold:
observations: 5
confidence: 0.80
sections:
- overview
- concepts
- exercises
Example 2: Adding event subscriptions
event_subscriptions:
- event: "minecraft.redstone.circuit.built"
handler: record_pattern
- event: "minecraft.redstone.debug.triggered"
handler: record_error_pattern

