Skip to content

Usage model

Operational roles in Merlon Docs.

Agents should load the parent Agent Skill (/merlon-docs, installed in your agent’s skills directory via npx skills add) first. That skill teaches how to query with smallest context — discover one shard at a time — without loading entire guides. See Agent Skill.

Actor Reads Writes Must run
Author (human/agent) shards, Agent Skill shards, manifest check before PR
CI config — check --require-lint when peers configured
Agent (context consumer) Agent Skill, single shards (rg/IDE) — —

Authors can also run merlon prose locally to see Vale findings before CI does.

Shards are source of truth; compiled files are generated.

One guide that compiles to guide.md and passes check. Install the skill with its install command.

Multi-guide compileOrder, publish outputs (compile.outputFile).

Above plus the four-tier guide layout (features / client / developer / glossary), the merlon-docs skill’s workflows.

Alternatives and adoption says what Merlon Docs adds beside site generators and host rules files. Scope and positioning explains why MCP sits on top of Merlon Docs instead of replacing it.

  1. When the parent Agent Skill (/merlon-docs) is available, load it with the task; the skill picks a workflow and runs intake for WORK_ITEM, as Skill workflows says
  2. Find the shard with host search and read one shard, as refs says
  3. Rely on merlon check for broken # cross-links (optionally inspect merlon refs-list)

Read docs/skills.md for the skill and its workflow index.