Skip to main content

What Coverage Indicators Are

Every section in every topic article carries a coverage tag that records how many source files contributed to that section and how well-synthesized it is. This lets Claude — and you — make a fast, informed decision: read the wiki section and move on, or go deeper into raw files.

The Three Levels

High

5 or more sourcesThe section is well-synthesized and recently compiled. Trust it directly and skip the raw files.

Medium

2–4 sourcesGood overview, but may miss detail on granular questions. Check raw sources if you need specifics.

Low

0–1 sourcesSparse data. Read the raw sources listed in that section directly rather than relying on the wiki.

What Coverage Tags Look Like

Coverage tags appear inline in each section heading:
The tag format is always: [coverage: {level} -- {N} sources]
Coverage varies by section within the same article. A topic might have a high-coverage Summary (synthesized from 15 meetings) and a low-coverage Gotchas section (only one source mentions it). Read accordingly.

How Your Agent Should Use Coverage

The most effective pattern is to have Claude apply coverage-aware reading automatically. You do this by adding a knowledge base block to your AGENTS.md or CLAUDE.md:
Once this is in your AGENTS.md, Claude uses coverage indicators automatically every session — no special commands needed.
The Sources section at the bottom of every article lists backlinks to every raw file that contributed. For medium and low coverage sections, these links tell Claude exactly which files to read — no broad scanning required.

The Practical Tradeoff

Coverage indicators give you the speed of a compiled wiki without sacrificing accuracy on sections that need depth. In practice, most heavily-used topics accumulate high coverage quickly. The [coverage: low] tag is most useful for newly discovered topics or sections that haven’t had many sources classified into them yet.
Run /wiki-lint to get a report of all low-coverage sections across the wiki. This shows you where the compiled knowledge is thin and which raw files you might want to add more detail from.