Plugin Layout

This repository is intentionally small. The plugin bundle is concentrated under plugins/specark/, while repo-root files handle distribution and top-level documentation.

Top-level structure

.agents/plugins/marketplace.json
plugins/specark/
  .codex-plugin/plugin.json
  references/
  scripts/
  skills/
README.md
CHANGELOG.md
docs/

Key directories

plugins/specark/skills/

Contains the callable skill folders. Each skill includes:

• SKILL.md
• agents/openai.yaml

The callable skills are intentionally thin wrappers. They point either to upstream source command text or to SpecArk-authored local command text.

plugins/specark/references/

Contains shared reference material that multiple skills may rely on, including:

• orchestration rules
• REASONS Canvas terminology
• upstream source command files in source-commands/
• SpecArk local command files in local-commands/
• authoring guidance for future skills

plugins/specark/scripts/

Contains helper scripts used by skills or by repository validation. These exist to keep repetitive or deterministic logic out of skill prose.

Design principle

The bundle follows progressive disclosure:

• lightweight metadata triggers a skill
• SKILL.md names the authoritative command file and maps repository terms
• reference files and scripts carry deeper detail only when needed

This keeps callable skills small without losing rigor.