feature-request-claude-code-link-resolution

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Claude Code already auto-resolves code imports — writing "from foo import bar" in Python triggers reading foo.py automatically. But any reference syntax in markdown files is completely ignored. Whether you use wikilinks, standard markdown links, or bare relative paths, Claude Code treats them all as dead text and never follows them.

When AI instructions are split across multiple files — which becomes necessary beyond roughly 50 lines to prevent attention dilution — the entire instruction graph collapses. Claude only sees the current file. Every downstream file becomes invisible, no matter how carefully the reference chain was designed.

Proposed Solution

When Claude Code reads a markdown file containing explicit file references, it should follow them, exactly as it follows import statements in code. The syntax doesn't matter — wikilinks, standard markdown links, or bare relative paths could all work. The behavior is straightforward: read a file, find a reference, read the target file, find more references, continue. Same mechanism as code imports applied to document references.

Alternative Solutions

The first alternative is a manual bridge file, which is what I use now. It requires 110 lines of hardcoded relative paths, breaks whenever a file moves or is renamed, and duplicates information already encoded in the links themselves.

The second alternative is embedding all instructions in a single file. This produces 400-plus line documents where rules are skimmed, missed, or forgotten — exactly the problem that splitting into modules was meant to solve.

The third alternative is an MCP server that intercepts file reads and resolves links. This is too heavy for most users and shouldn't require infrastructure for something as basic as following a link.

Priority

Medium - Would be very helpful

Feature Category

Developer tools/SDK

Use Case Example

My Obsidian vault uses CLAUDE.md files to define AI behavior across roughly 30 files connected by wikilinks. Starting a new project involves a 7-step workflow: a routing table file points to a workflow map, which points to seven step files, each of which references domain-specific skills, checklists, and a shared toolbox. The full instruction graph is deep and deliberately modular.

Without reference resolution, Claude Code reads the first file, sees a wikilink as plain text, and stops. The other 29 files are unreachable. The entire instruction architecture is invisible.

With reference resolution, Claude Code would traverse the reference chain, and the full instruction graph would activate automatically — just as it does when following Python imports through a codebase.

I currently maintain a 110-line bridge file at the project level that manually lists every path, essentially reimplementing link resolution in prose.

Additional Context

I'm using Claude Code v2.1.158 via the VS Code extension, with an Obsidian vault as the workspace. This is not an Obsidian-specific request. Any team that scales AI instructions beyond a single file hits the same wall, regardless of what link syntax they use. Wikilinks happen to be the most common syntax today because Obsidian popularized them, but the underlying need is universal: document references should resolve the same way code imports do. Claude Code already understands the concept of "go read that file" — it just doesn't apply it to markdown yet.