Learn Merlon Docs
New to Merlon Docs? Work through this path from top to bottom, or jump to the step you need. Each step ends with somewhere to go next.
Start here
Section titled “Start here”Merlon Docs comes in three parts. Part 1 is the Agent Skill (merlon-docs, which teaches your coding agent how to read, write, and review shards). Part 2 is the skill’s workflows, one for each kind of task. Part 3 is the CLI (merlon compile and merlon check, which build guides and validate links in CI). Read the Agent Skill directly first, or start with why documentation reliability matters.
1 · Understand the problem
Section titled “1 · Understand the problem”Software failures cost hundreds of billions to trillions of dollars a year, and documentation is one of the causes a team can act on. When agents write much of the change, docs also break in a few familiar, structural ways.
2 · Understand the ideas
Section titled “2 · Understand the ideas”Start with the shard, then read how guides, the check and the skill build on it.
3 · Try it
Section titled “3 · Try it”Install the skill in your agent, or the CLI in your repository, and run a compile. The quick start takes a repository from no shards to a passing check.
npx skills add Merlon-Docs/merlon --skill merlon-docs4 · Measure it
Section titled “4 · Measure it”How much context does an agent load? This repository measures a whole compiled guide against one shard. The figures are rough estimates. Every model and system tokenizes differently, so compare the sizes with each other, not with a bill.
| What the agent reads | Relative size | Tokens (estimate) |
|---|---|---|
| Features guide and glossarythe monolith, docs/_build/guides.md | 85,071 | |
| Developer guideDEVELOPERS.md | 41,301 | |
| CLI guidepackages/merlon-docs-cli/README.md | 17,051 | |
| Largest shardone file | 4,322 | |
| 90th percentile shardone file | 2,092 | |
| Average shardmean of 148 shards | 980 | |
| Median shardone file | 807 |
bb3d832. Counted with o200k_base from the open-source gpt-tokenizer. Every model and system tokenizes differently, so read these as rough, relative sizes, not costs. Benchmarks says how to re-measure with pnpm bench:context-size and another tokenizer or a counter of your own.After pnpm build && pnpm docs:compile:repo, this command counts this repository’s docs with cl100k_base. With --no-write, it prints the table and doesn’t write a CSV:
pnpm bench:context-size --tokenizer cl100k_base --no-write5 · Compare to what you know
Section titled “5 · Compare to what you know”Already keep a long README, a wiki or an llms.txt file? See where Merlon Docs agrees, where it differs, and
how it relates to the Model Context Protocol.
Where to next
Section titled “Where to next”Past the basics? Pick a direction.