For AI agents

The whole site is plain markdown underneath. An agent should never read all of it — read the map, pick a skill, then fetch only the notes the task needs.

1. Read the map

/llms.txt lists every skill and every note with a one-line summary and a link to its raw markdown.

2. Load a skill

Each skill (/skills/<name>.md) is a short playbook plus a note map saying which note to read when. Its file paths look like ${CLAUDE_PLUGIN_ROOT}/wiki/mobile-apps/aso/<slug>.md — on the web that is/wiki/<slug>.md.

3. Fetch notes

/wiki/<slug>.md returns the note with YAML frontmatter (summary, related,sources, updated). Citations look like yt:<video id> @ mm:ss — Speaker, i.e.https://www.youtube.com/watch?v=<id>&t=<seconds>.

4. Follow links

[[slug]] links and related: point to other notes. The full graph (nodes with summaries and backlinks, edges) is at/graph.json.

Claude Code

The same content is packaged as a Claude Code plugin, where skills load automatically by their description:

/plugin marketplace add <path-or-repo>
/plugin install subida-kb@subida

How to use the knowledge well

• Numbers are usually from single tests or one agency's clients — present them as "in one test", not as laws.
• Check updated and source dates: platform rules (Apple, Meta, Google) change.
• Each skill lists its gaps. If a question falls there, say so instead of improvising.