System Documentation Rules

Indexes and Compendiums

• Player-facing rules live under system/player/; GM-only construction and conversion guidance lives under system/gm/.
• Canonical machine-readable records, review ledgers, and generated indexes live under system/data/; do not expose them as peer rule pages.
• Retired system views and superseded generated caches live under archive/system/, with redirects from their former public URLs.

• The system exposes exactly eight compendiums: approved and candidate collections for Items, Monsters, Skills, and Effect Mechanics.
• Spells, programs, embodiments, equipment upgrades, and effect applications are represented inside Items, Skills, or Effect Mechanics; never create separate system compendiums for them.
• Candidate compendiums are visible working collections but are not playable or approved rules. Source-review caches remain evidence and are not promoted into candidate compendiums automatically.
• Moving a record between an approved and candidate pair requires an explicit editorial decision. Preserve its ID and content; change only its status and collection unless the user separately approves content edits.
• Review checkboxes and comments write only to the paired *_feedback.json file; feedback never changes compendium membership automatically.
• Follow System Compendium Contract for provenance, normalization, merging, naming, and approval requirements.

• Every documentation page has exactly one role: index or content. Never create a hybrid page.
• An index is navigation, not a container for item rules.
• This rule applies to every compendium, including monsters, NPCs, fauna, flora, items, spells, skills, maps, senses, biomes, and campaign records.
• Do not inline complete or partial record rules, stat blocks, effects, descriptions, procedures, or examples in an index page.
• Every named compendium record must have its own canonical page.
• Index rows may contain only navigation metadata needed to find the record, such as name, category, type, or governing stat.
• Link every index entry directly to the record's page; do not link to an anchor containing an inline record on the index.
• A content page documents one named record or one rules topic. It may link back to indexes but must not become an index for peer records.
• Structured JSON is the source of truth when a compendium provides it. Generated Markdown pages are committed caches for the wiki, CLI tools, and agents.
• Render both the index and individual record pages from the same JSON so their names, paths, and metadata cannot drift.