| name | wiki-seed |
| description | One-time bootstrap of the Roblox/Luau wiki from existing repo content. Extracts Luau/Roblox knowledge from .claude/agents/, .claude/rules/, and .claude/docs/ and generates initial wiki pages with proper frontmatter, cross-references, and ownership. Run ONCE at project setup; subsequent updates use /wiki-ingest. |
| disable-model-invocation | true |
/wiki-seed — Wiki Bootstrap
Delegate to: wiki-curator
When to Use
- Exactly once, when first setting up the wiki for a project that already has FoG-Roblox-Studio-Command content in
.claude/
- Never run a second time — use
/wiki-ingest for ongoing updates
If wiki/services/ already contains files, refuse to run and tell the user "Wiki already seeded. Use /wiki-ingest to add new content or /wiki-update to edit."
Prerequisite Files
Before running, verify these exist:
wiki/SCHEMA.md — the schema (read it first)wiki/README.md — the entry point- Empty
wiki/services/, wiki/concepts/, wiki/luau/, wiki/anti-patterns/, wiki/exploits/, wiki/performance/, wiki/monetization/, wiki/studio/, wiki/patterns/
Workflow
1. Read the Schema
Load wiki/SCHEMA.md fully. Every page created must comply with it.
2. Inventory Source Material
Read these existing repo files (in order):
Prescriptive rules (for anti-patterns and required patterns):
.claude/rules/server-scripts.md.claude/rules/client-scripts.md.claude/rules/shared-modules.md.claude/rules/datastores.md.claude/rules/remotes.md.claude/rules/ui-code.md.claude/rules/gameplay-systems.md.claude/rules/design-docs.md.claude/rules/tests.md.claude/rules/prototypes.md.claude/rules/config-data.md
Architecture and language docs:
.claude/docs/roblox-architecture-guide.md.claude/docs/luau-style-guide.md.claude/docs/coding-standards.md
Agent prompts (for Roblox-specific knowledge sections):
.claude/agents/datastore-architect.md.claude/agents/remotes-networking-specialist.md.claude/agents/exploit-security-specialist.md.claude/agents/analytics-retention-specialist.md.claude/agents/live-ops-specialist.md.claude/agents/monetization-lead.md.claude/agents/performance-analyst.md.claude/agents/technical-director.md.claude/agents/lead-programmer.md.claude/agents/luau-gameplay-programmer.md.claude/agents/luau-systems-programmer.md.claude/agents/ui-programmer.md.claude/agents/art-director.md.claude/agents/audio-director.md.claude/agents/sound-designer.md.claude/agents/level-designer.md.claude/agents/technical-artist.md.claude/agents/roblox-studio-specialist.md.claude/agents/devops-engineer.md
3. Extract Knowledge by Category
Scan all sources and extract (don't duplicate across categories):
Services (each becomes a wiki/services/<ClassName>.md page):
Every Roblox class mentioned by name. Priority order: DataStoreService, MemoryStoreService, OrderedDataStore, RemoteEvent, RemoteFunction, UnreliableRemoteEvent, MarketplaceService, Players, TextChatService, MessagingService, TweenService, TeleportService, HttpService, SoundService, CollectionService, ContextActionService, UserInputService, ProximityPrompt, RunService, Lighting, Humanoid, and every other class-name referenced in agent prompts.
Concepts (wiki/concepts/*.md):
- session-locking
- server-authority
- client-server-split
- bind-to-close
- schema-versioning
- trove-maid-cleanup
- signal-pattern
- service-pattern
- rate-limiting
- streaming-enabled
- module-lazy-loading
- atomic-trading
- cross-server-events
- feature-flags
- code-redemption
- ftue-design
- core-loop
Luau features (wiki/luau/*.md):
- type-annotations
- strict-vs-nonstrict
- task-library
- export-type
- pcall-xpcall
- table-library
- string-library
- math-library
- buffer-type
- generic-types
- coroutines
- metatables
- require-pattern
- module-scripts
Anti-patterns (wiki/anti-patterns/*.md):
- deprecated-wait
- deprecated-spawn
- deprecated-delay
- client-trust
- magic-numbers
- no-pcall
- no-session-lock
- instance-in-remote
- client-to-server-remote-function
- player-name-as-key
- missing-schema-version
- print-in-production
- no-rate-limit
- unvalidated-remote-args
- string-concat-in-loop
- direct-cross-system-coupling
Exploits (wiki/exploits/*.md):
- speed-hack
- teleport-hack
- fly-hack
- noclip
- remote-spam
- argument-spoofing
- item-duplication
- negative-purchase
- transaction-replay
- localscript-injection
- memory-editing
- bind-to-close-skip
- session-hijack
Performance (wiki/performance/*.md):
- heartbeat-budget
- client-fps-targets
- server-memory-budget
- network-bandwidth-budget
- microprofiler
- developer-console-stats
- script-profiler
- object-pooling
- throttling-pattern
- caching-references
- string-concat-tableconcat
- overdraw-transparency
- particle-budget
- dynamic-lights-budget
- streaming-performance
- datastore-budget-management
Monetization (wiki/monetization/*.md):
- game-pass
- dev-product
- process-receipt-idempotency
- premium-benefits
- engagement-based-payouts
- robux-price-tiers
- developer-exchange
- ethical-monetization
- gamepass-detection
- purchase-funnel
Studio (wiki/studio/*.md):
- rojo-mapping
- argon-mapping
- wally-packages
- collection-service-tags
- attributes
- play-solo-team-test
- device-emulator
- studio-plugins
- open-cloud-api
- rbxl-vs-rbxlx
Patterns (wiki/patterns/*.md):
- daily-rewards
- code-redemption-system
- quest-system
- inventory-pattern
- trading-system
- party-system
- friends-system
- leaderboard-pattern
- onboarding-tutorial
4. Create Pages
For each extracted topic:
-
Write a complete page (status: complete) IF:
- The existing repo content has enough material for a substantial page (≥3 paragraphs of relevant content + code examples)
- This is a core topic the rules or agents deeply cover
-
Write a stub (status: stub) IF:
- The topic is only briefly mentioned in existing content
- It deserves its own page but needs more source material
Target distribution for medium-depth seed:
- ~35-45 complete pages (the rules and architecture guide provide enough content for these)
- ~80-90 stub pages (so nothing is a dangling reference, but fleshing out waits for
/wiki-ingest from the raw sources layer)
5. Cross-Reference
After creating all pages, do a second pass:
- Add
[[wikilinks]] in every Related section
- Ensure bidirectional links (if A's Related includes B, B's Related should include A)
- Link from concept pages to service pages they use, and back
- Link from anti-pattern pages to the rule file that enforces the rule (use markdown link for rules, wikilink for wiki pages)
- Link from exploit pages to the concepts that defend against them
6. Write wiki/index.md
Full content catalog, grouped by category, with one-line summary and status per page.
7. Write wiki/log.md
First entry:
# Wiki Log
## [YYYY-MM-DD] seed | Initial wiki bootstrap
- Seeded from: .claude/agents/ (33 files), .claude/rules/ (11 files), .claude/docs/ (3 files)
- Pages created: <N total>
- Complete: <N>
- Stub: <N>
- Categories:
- services: <N>
- concepts: <N>
- luau: <N>
- anti-patterns: <N>
- exploits: <N>
- performance: <N>
- monetization: <N>
- studio: <N>
- patterns: <N>
- Owners assigned per schema ownership table
- Cross-references: <N total wikilinks>
- Next: /wiki-ingest the raw sources captured by research agents
8. Final Check
Run /wiki-lint automatically after seeding to verify:
- All frontmatter valid
- No broken links
- All pages in index.md
- No unexpected orphans
Report the lint result to the user alongside the seed summary.
Output to User
After completion, report:
- Total pages created (complete + stub)
- Breakdown by category
- Any skipped topics and why
- Lint report summary
- Recommended next step: "Run /wiki-ingest on raw sources captured by research agents"
Anti-Patterns for the Seeder
- ❌ Do not write pages longer than the source material justifies (better to stub and extend later)
- ❌ Do not invent Roblox APIs or behaviors not in the source material
- ❌ Do not create pages for topics you couldn't find any source content on (defer until a source arrives)
- ❌ Do not overwrite existing wiki files if any somehow exist (fail instead — the user needs to clear them)
- ❌ Do not skip the index.md or log.md — those are required outputs
- ❌ Do not create circular Related links that don't actually relate (only link pages that are genuinely connected)
