book-game
Skill40 starsChanged 2 months agoApache-2.0View on GitHub
Game narrative craft reference: branching architecture, quest design, dialogue systems, lore bible, player agency, and game-writing AI tells. Read by all agents on game scenarios, visual novels, and interactive fiction.
In this repository23 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- dashboard/CLAUDE.mdCLAUDE.md
- .cursor/rules/velith-editing.mdcCursor rule
- .cursor/rules/velith-genres.mdcCursor rule
- .cursor/rules/velith-pipeline.mdcCursor rule
- book-academicSkill
- book-draftSkill
- book-editSkill
- book-fictionSkill
- book-genre-creatorSkill
- book-ideationSkill
- book-illustrateSkill
- book-initSkill
- book-nonfictionSkill
- book-outlineSkill
- book-poetrySkill
- book-publishSkill
- book-screenplaySkill
- book-statusSkill
- book-technicalSkill
- book-visualsSkill
- loomSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Critical path (60%): must work if the player does nothing optional.
- Distinctive: line 45
Every choice offering good / neutral / evil.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Game Narrative
The reader is a player. They act, and the story must answer. A game script that reads well but ignores the player's hands is a screenplay with buttons.
Architecture
- Critical path (60%): must work if the player does nothing optional. Every beat of the main story lands regardless of side content.
- Side content (30%): enriches, never gates. Side quests reward with understanding of the world or characters, not only items.
- Hidden (10%): for players who look. Rewards curiosity.
- Branch depth ≤ 4 before convergence. Convergence types: full (all paths merge), partial (state carries), butterfly (small early choice, large late consequence; use once or twice per game, telegraph nothing).
- Choice types: binary, multi-way, conditional (flag/item), timed. A choice with no consequence is a menu, not a choice. A choice where one option is obviously correct is not a choice.
- Endings: each ending is earned by a pattern of choices, not one final selection. The player should be able to explain why they got their ending.
Quest design
Per quest: ID, title, hook (why the player cares), prerequisites (flags, items, relationships), steps (3-7, each verifiable by the game), branch points, resolution, rewards, flags set. Quest chains: linear, hub, parallel.
The hook is narrative, not "collect five." The reward includes a change in the world the player can see.
Dialogue system
- Node: speaker, line, emotion tag, next-node links, conditions.
- Player responses ≤ 3 per node; each is a stance (investigate, push, withdraw), not a paraphrase of the same thing.
- NPC state (trust, fear, knowledge) affects available lines. Track in the bible.
- Barks and ambient lines carry world texture; write them with the same care as cutscene dialogue.
- Lines are short. Players skip. The most important line in a conversation is the one that changes state, and it should be unmissable.
- Localization: avoid idioms that do not travel; keep string lengths sane for UI.
Lore bible
Entries per location, character, faction, item, event, concept, rule. Fields: name, description, connections, mechanical relevance, narrative role, first appearance, flags. Consistency check before finalizing: every named thing in a quest or line exists in the bible. World rules (magic, tech, politics) are stated once and never broken without a story reason.
Character across routes
Core identity is constant; route-specific development varies. A companion who is warm on one route and cold on another needs a reason the player can find. Personal quests unlock on trust thresholds. Companions comment on the player's choices; silence is also a comment.
Game-writing tells
- NPCs who explain the world to a player character who lives in it.
- Every choice offering good / neutral / evil.
- Quest text that restates the objective marker.
- Lore dumps in item descriptions that nobody reads.
- Villains monologuing at the boss door.
- Companions who approve of everything.
- Endings differentiated by one final dialogue choice.
- Uniform dialogue length across NPCs.
Format
Markdown with YAML frontmatter per quest and per dialogue tree. Branching maps as Mermaid flowcharts. Dialogue as structured blocks:
### NODE guard_gate_01
**Speaker:** GUARD · **Emotion:** wary · **Condition:** !has_pass
"State your business."
- [investigate] "What happened here?" → guard_gate_02
- [push] "Let me through." → guard_gate_03 (sets: guard_hostile)
- [withdraw] (leave) → ENDDeliverables per outline: overview, world bible, character bible, main quest line, side quests, dialogue trees, cutscene scripts, branching map, flag table.
Validation (architect scoring for game)
Critical path works standalone; branch depth within limit with defined convergence; every choice has consequence; endings earned by patterns; every named entity in the bible; flag table complete and no orphan flags; quest steps verifiable.
Visuals
The scenario document needs: world map and region maps (figure-engineer base layers from the lore bible, stylized by the illustrator), branching maps (Mermaid flowcharts from the quest data, never drawn by hand), character sheets with mechanical annotations, location key art, item and UI mock references. Character and location sheets are locked before scene art. Consistency is not optional: the art bible's constants become the studio's reference. All through /book-visuals.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

