Source Notes

These notes explain why the template looks like an operational stack pack instead of a conventional learn topic.

What Shaped The Template

Source area	Influence
Lab stack research	Stack packs should answer shipping questions, not reproduce framework docs
Context injection research	Keep README small; move details into focused files with raw recovery paths
Agent skill ecosystems	Use progressive disclosure, trigger-rich skill descriptions, support files, and eval loops only when useful
Copilot plugin directories	Capability packs can eventually bundle skills, hooks, MCP, agents, and LSP config
Microsoft skill tests	Acceptance criteria and scenarios can catch generated-code mistakes
gstack	Quality gates should be tiered, command-backed, and explicit about paid/manual checks

Refresh this template when any of these change:

Agent Skills specification frontmatter, directory shape, or validation behavior.
OpenCode, Copilot, Claude Code, or Codex skill discovery paths.
GitHub CLI gh skill preview, install, update, pinning, or publish validation behavior.
Research Lab rules for _repos/, source tracking, learn topics, or stack docs.
A concrete stack pack exposes a missing required file or recurring failure mode.

The template intentionally does not include concrete Android, Tauri, Windows, Expo, Flutter, PWA, or local-first commands.
The OpenAI openai/plugins repo was not fetched for this run; plugin distribution details should be refreshed before building a real plugin bundle.
Local-docs had several relevant repos in queued or indexing status, so direct local file reads grounded the most important claims.
This template does not authorize package installs, repo clones, builds, tests, or _repos/ mutation.