adventure-writing

name

adventure-writing

description

Use when designing, drafting, or revising a tabletop RPG adventure, scenario, campaign chapter, or sandbox opener for any system. System-agnostic. Covers the full architecture: cold open, historical/situational frame, faction wants, NPC casting, rumor seeds, consequence tracks, scene-by-scene branching grammar, dilemma design, distributed clues, combat-encounter blocks, multi-ending reckonings, sandbox exits, and pacing rhythms. Built from the design DNA of robust, replayable, choice-driven adventures that survive contact with players. Triggered by requests to "write an adventure", "design a scenario", "outline a campaign chapter", "draft a sandbox opener", "build an introductory module", or to revise/audit an existing adventure for branching, pacing, or robustness.

Adventure Writing — A Bible for Branching, Replayable Scenarios

This skill exists because most adventures collapse into railroads, puzzle-locked clues, or scripted heroics the moment players touch them. The published modules that survive contact with the table share a common architecture: conflicting factions with stated wants, distributed truth, consequence tracks, dilemmas without clean answers, and endings conditional on player history.

This bible teaches that architecture, system-agnostic. It works for d20, PbtA, FitD, year zero, OSR, GUMSHOE, Free League, or custom systems. It does not teach mechanics; it teaches the shape of an adventure that resists collapse.

The skill is long because the job is large. Read this file first. Then load the modules you need.

Published register

This skill distinguishes between design language and published adventure language.

• Design language belongs in the skill files, learnings, outlines, and private notes. It can use technical shorthand such as lever menu, consequence ledger, truth-survives table, or other internal labels that help the author think.
• Published adventure language belongs in the manuscript the GM will read at the table. It should sound like commercially published RPG material: plain, instructive, confident, and easy to run. Prefer routes, options, approaches, faction standing, consequences, timers, openings, scenes, and developments over workshop jargon.

Two rules follow from this:

1. Do not print internal craft terms in the adventure text unless they are already standard in commercial game writing. Arc, clock, and track are fine. Avoid terms like beat, lever, consequence ledger, toolkit, or payoff when a plain-language alternative will do the job.
2. GM guidance should read as usable module prose, not theory. Write "If the party opens in Baldy, bring in Eufemia's freight problem early" rather than "make the second beat the freight lever." The GM should feel instructed, not lectured on craft.

The non-negotiables

These rules are absolute. Violate them and the adventure becomes a railroad, a quiz, or a tour.

1. Every faction has a Want, a Method, and a View of the PCs. No faction is "the bad guys." Every faction acts on its own interest. The party is a variable, not the protagonist.

2. No single roll, NPC, or clue is load-bearing. Every important fact lives in at least three mouths, pockets, or places. Failure rerouts; it does not stall.

3. Every scene offers four or more genuine player branches. Comply, refuse, subvert, exploit, ignore. If a scene only tests "did they roll high enough", rewrite it.

4. Dilemmas precede dice. The hard moment in a scene is a moral or practical choice the players make as themselves. The roll resolves the cost, not the decision.

5. Consequences are tracked, not narrated. Adventures use a small number of named, system-agnostic counters (+1 / -1) that cash out at the ending. The numbers shift in fiction, not in handwave.

6. Every ending is conditional. The reckoning has at least four distinct end-states determined by the tracks, not by GM decree. "Bad endings" must still seed the next adventure.

7. The cold open shows tone, not exposition. Open on a sensory image of the world's specific weight. Save the briefing for the next section.

8. Every NPC has a stance toward the PCs. Not just "what they want" — what they think the PCs are (tools, threats, marks, witnesses, peers). That stance evolves.

9. Robustness is a feature. Every important NPC has a "what if they die early" branch. Every clue has a fallback carrier. The adventure must survive a TPK of any single NPC.

10. End on weight. The end of every scene is a cost, a decision, or an image — never a summary.

The workflow

Use this order when drafting a fresh adventure. Skipping a step produces the failure modes you will recognize in bad modules: plot holes, dead clues, single-point failures, scripted endings.

Phase 0 — Scope and Research

Do not draft a single scene until these two steps are complete.

0a. Set scope before drafting. Decide: one-shot / campaign chapter / mini-campaign. Set a word-count target and a session- count target. Scope failures (too big, too small) produce unpublishable drafts. See references/scope-and-format.md. 0b. Run the research pipeline. Answer six research questions about the world's political situation, economy, technology, displaced people, daily work, and recent history. Do not draft until all six can be answered in two sentences each. See references/research-pipeline.md.

Phase 1 — Design

1. Pick the structure type. Decide first: is this adventure a three-act chapter (the party drives toward a climax; pacing cohesion is primary) or a fixed-node sandbox (the party moves freely through a country of named locations, choosing what to care about; freedom of movement and order of operations is primary)? The two structures share most craft but diverge on cold open, prize design, time pressure, and the "what if they break it" section. See references/sandbox-structure.md for the full comparison and the four sandbox properties (independent levers, faction clocks, entry-point menu, shortcut doctrine). Most adventures are one or the other; some combine them (a chapter whose macro rhythm is three acts but whose scene layer is a sandbox). State the choice before drafting.
2. Pick a pressure, not a plot. State the central pressure in one sentence: "A railroad survey forces a wagon column across hunting ground." Pressure is what makes choices hurt.
3. Cast the factions. Three to six. Each gets one line for Want, one for Method, one for View of PCs. See references/factions-and-tracks.md.
4. Define the consequence tracks. Three or four named counters, one per major faction or stake. See references/factions-and-tracks.md.
5. Cast the named NPCs. Use the NPC template in references/npcs.md. Each entry: physical, defining quality, mannerism, speech register, view of PCs.
6. Write the cold open. A sensory paragraph in italics. Then the stakes paragraph. Then the historical/situational frame. See references/architecture.md and references/read-aloud.md.
7. Lay the arcs. Three acts: setup (breath), pressure (hammer), reckoning (cash-out). See references/pacing-and-arcs.md.
8. Design scenes as branches, not steps. Each scene: read- aloud, ground-truth, NPC stance, four+ branches, track shifts. See references/scene-design.md.
9. Distribute the truth. No fact lives in one place. Use the three-clue rule and the carrier-redundancy table. See references/clue-discipline.md.
10. Write the combat blocks. Enemies / area / terrain effects / how the fight ends in fiction. See references/combat-encounters.md.
11. Write the dilemmas. One genuine no-clean-option choice per arc. See references/dilemmas.md.
12. Write the multi-ending reckoning. Four+ end-states keyed to track totals. See references/endings-and-sandbox.md.
13. Add the failure-mode section. "If X dies early," "If party sides with the antagonist," "If they refuse the hook." See references/architecture.md.
14. Run the anti-railroad pass. Cut every "the players must," "the GM should ensure," and scripted outcome. See references/anti-railroad.md.
15. Add sandbox exits. What threads survive into the next chapter. See references/endings-and-sandbox.md.

Phase 3 — Playtest and Feedback Loop

Run these steps after every live session of the adventure.

15. Playtest. Run the adventure at a live table or arrange a controlled playtest session. Your job during play is to observe and take notes — not to explain the text or fill gaps. Gaps and stalls are data. See references/playtesting.md.
16. Collect structured feedback. Run the post-session debrief 12-24 hours after the game, not immediately after. Use the five-question form. See references/feedback-collection.md and templates/feedback-form.md.
17. Triage and file learnings. Sort feedback into three buckets: (A) structural fixes for this adventure, (B) author learnings to prevent future failures, (C) preference mismatches (no fix required). File bucket-B learnings in learnings/. Apply bucket-A fixes before publishing. See learnings/README.md.

The library

references/ — how to design

• references/architecture.md — the macro shape. Cold open, stakes paragraph, frame, shape-of-it, cast, rumors, arcs, reckoning, pacing notes, GM tone notes. The skeleton every chapter shares.
• references/scene-design.md — the meso shape. The read-aloud / ground-truth / NPC behavior / four-branches / track-shift pattern. Camp scenes vs. side events vs. set-pieces.
• references/npcs.md — the cast template. Body, defining quality, mannerism, speech register, view of the PCs. Why every NPC needs a stance. How dialogue functions as fingerprint. Robustness clauses.
• references/factions-and-tracks.md — the consequence ledger. How to pick three or four named tracks. How factions act on high vs. low scores. The Want / Method / View triplet.
• references/clue-discipline.md — distributed truth. The three-clue rule. The carrier-redundancy table. How to spread a fact across NPCs, objects, and environments so no roll kills the chapter.
• references/pacing-and-arcs.md — three-act rhythm. Breath, hammer, reckoning. Camp scenes for downbeats. Side events for pressure injection. How to know when the middle is sagging.
• references/dilemmas.md — choice design. The no-clean-option principle. Mutually exclusive goods. How to make every road cost. How to keep dilemmas off the dice.
• references/combat-encounters.md — the combat block. Enemies, area, terrain effects, what changes when the fight ends. Where combat lives in the arc. Combat as consequence-amplifier, not as plot beat.
• references/endings-and-sandbox.md — the reckoning. Four+ end-states keyed to tracks. Ending as launchpad. Sandbox exits. What the party carries out. What hunts them now.
• references/historical-grounding.md — real-world weight. Historical-note inserts. Period anchors. Why grounding makes fantasy and sci-fi heavier too. The research register.
• references/anti-railroad.md — the kill list. Phrases, patterns, and structural moves that turn an adventure into a rail. Detection cues and rewrites.
• references/read-aloud.md — sensory openers. What goes in italics, what stays in the GM voice. Cadence, smell, light, sound, weather. The kill list for read-aloud prose.

references/ — Phase 0 and Phase 3 process files

• references/scope-and-format.md — scope calibration. One-shot vs. campaign chapter vs. mini-campaign. Word count and session- count targets. Format decisions (pre-gens, handouts, maps). The minimum viable adventure.
• references/research-pipeline.md — the pre-draft research workflow. The six research questions. Source hierarchy. When to stop researching. The research register format.
• references/playtesting.md — controlled playtest protocol. Observer role, note-taking shorthand, intervention threshold, blind playtest procedure.
• references/feedback-collection.md — post-session debrief. Timing, the five questions, feedback triage, protecting the author.

references/ — expanded design domains

• references/investigation-structure.md — mystery and investigation adventure design. Crime-scene reconstruction, suspect webs, clue vs. conclusion distinction, escalating timelines, open-loop endings.
• references/node-based-design.md — node-based scenario design. Node types, designing backward, floating clues, chokepoints, exit design, relationship maps.
• references/sandbox-structure.md — the sandbox as a fourth adventure shape, distinct from the three-act chapter. The fixed-node sandbox, the three sandbox substyles (social, pointcrawl, hexcrawl — what the nodes are made of and therefore what the party's primary engagement is), the lever menu, faction clocks that advance on time, the entry-point menu, shortcut doctrine (true consequences, never resets), the sandbox reckoning, and when to pick sandbox vs. chapter. Load this before designing any adventure whose pleasure is the party choosing what to care about in a rich situation.
• references/levers-and-fallbacks.md — the lever menu: many sufficient routes to the reckoning, none mandatory. Lever design rules, the cost-distinctness rule, fallbacks for missed routes, and the truth-survives-the-holder table. The action-layer companion to clue-discipline's three-clue rule. Contains the language-register rule: design concepts stay in design files; adventure text uses consumer-facing language (route, not lever; faction standing, not consequence ledger; scene sequence, not beat map). Load for any adventure, chapter or sandbox, that must survive the party refusing the obvious path.
• references/horror-design.md — dread and tension architecture. Gothic, cosmic, and survival horror modes. The unseen principle, the reveal economy, horror pacing, safety toolkit.
• references/antagonist-design.md — deep villain construction. Motivation archaeology, mirror-protagonist principle, antagonist competence, hierarchy, perspective paragraph.
• references/location-design.md — keyed location design. Entry format, environmental storytelling, three-zone rule, secret placement, the living location, mapless design.
• references/social-encounters.md — social scene design. Pivot points, stakes, pressure, social dilemmas, interrogation scenes, anti-patterns.
• references/clocks-and-timers.md — countdown mechanics. Clock types, visible vs. hidden clocks, triggers, cash-out, anti-patterns.
• references/travel-and-wilderness.md — overland travel, pointcrawl, hexcrawl, landmark navigation, encounter tables, journey compression, terrain and weather.

templates/ — drop-in scaffolds

• templates/chapter-skeleton.md — a blank chapter scaffold with every section header in order. Copy and fill.
• templates/investigation-skeleton.md — blank scaffold for mystery and investigation adventures. Organized around the crime, suspect web, node map, and resolution conditions.
• templates/one-shot-skeleton.md — compressed scaffold for self-contained one-shots. Single-session pacing, streamlined cast, two-to-three end-states.
• templates/feedback-form.md — post-session feedback questionnaire. Five player questions plus two GM questions. Print or share digitally.

learnings/ — table-sourced knowledge

• learnings/README.md — the learnings folder system. Format for individual entries, naming convention, how learnings graduate into reference files.
• Adventure-specific learnings files are created here after each playtest. Format: learnings/[adventure-slug].md.

examples/ — annotated exemplars

• examples/annotated-yellowstone.md — a chapter from a shipped adventure (the Yellowstone Line, a three-act chapter), annotated to show every pattern in this bible at work. Read this before drafting your first chapter with the skill.
• examples/annotated-quiet-title.md — a sandbox from a shipped adventure (Quiet Title, a fixed-node sandbox), annotated to show the sandbox-specific patterns at work: entry-point menu, lever menu, faction clocks, shortcut doctrine, truth-survives table, split-party design. Read this alongside the Yellowstone exemplar; the two are a pair showing the same craft applied to different structures.
• examples/comparative-patterns.md — ten design decisions shown side by side across both adventures (cold open, central prize, time pressure, "what if they break it," cast distribution, reckoning, historical grounding, consequence ledger, robustness, sensory registers). The agnostic tour: the architecture is portable, the shape serves the situation. Read when deciding chapter vs. sandbox, or when a pattern feels wrong and you suspect a structure mismatch.

How to use this bible

1. Always read this file first. It contains the non-negotiables and the workflow. Everything else supports.
2. For a fresh adventure, run Phase 0 first: set scope (scope-and-format.md), then run the research pipeline (research-pipeline.md). Then load architecture.md, factions-and-tracks.md, and npcs.md before drafting any prose. Then scene-design.md and clue-discipline.md for the body. Then endings-and-sandbox.md for the close.
3. For a scene-level draft, load scene-design.md, read-aloud.md, and dilemmas.md.
4. For a combat set-piece, load combat-encounters.md and scene-design.md.
5. For an audit or revision pass, load anti-railroad.md and clue-discipline.md. Run the kill list.
6. For a historical, sci-fi, or fantasy adventure that needs weight, load historical-grounding.md. The patterns are identical; only the source material changes.
7. For an investigation or mystery adventure, load investigation-structure.md and node-based-design.md. Use templates/investigation-skeleton.md for the scaffold.
8. For a horror adventure, load horror-design.md. Adjust pacing and read-aloud registers accordingly.
9. For a one-shot, load scope-and-format.md first. Use templates/one-shot-skeleton.md for the scaffold.
10. For a villain or antagonist with depth, load antagonist-design.md. Write the antagonist's perspective paragraph before drafting Arc I.
11. For travel sequences or wilderness adventures, load travel-and-wilderness.md and location-design.md.
12. For a sandbox adventure (the party moves freely through a country of named locations, choosing what to care about), load sandbox-structure.md and levers-and-fallbacks.md before drafting. Pull exemplars from examples/annotated-quiet-title.md. Read examples/comparative-patterns.md to keep the sandbox patterns distinct from chapter patterns.
13. Pull two or three exemplars from examples/annotated-yellowstone.md (for a chapter) or examples/annotated-quiet-title.md (for a sandbox) before drafting any section. Match the cadence.
14. Draft, then run the anti-railroad pass. Cut anything it catches. Rewrite.
15. End on weight. The last beat of every section is a cost, a decision, or an image — never a summary.
16. After every live session, run Phase 3 (steps 15-17). Observe without explaining. Debrief 24 hours later. File learnings before starting the next adventure.
17. For an adventure with difficult content, declare content types in the GM notes and recommend a session zero. See the Safety and content discipline section above. For horror specifically, the safety toolkit in horror-design.md is mandatory, not optional.

The acid tests

Before you call an adventure done, run all six.

1. The four-branch test. Open a random scene. Can you name four genuine player responses, each with a different downstream effect? If not, redesign.
2. The dead-NPC test. Pick the most important NPC. Kill them in the first scene they appear. Does the chapter still reach a reckoning? If not, redistribute their function.
3. The single-roll test. Search the chapter for any moment where a missed roll halts the adventure. There should be none. Every gate has at least three keys.
4. The four-ending test. Read your reckoning. Can a careful reader predict the end-state from the track totals without GM intervention? If not, the conditions are fuzzy.
5. The cold-open test. Read the first paragraph aloud. Does it teach the world's specific weight in three sentences, or does it summarize the plot? If the latter, rewrite.
6. The blind-GM test. Hand a single scene to a GM who has never seen the adventure. Can they run it from the text alone without asking you a question? If they need clarification, the scene's instructions, NPC entries, or exit descriptions are underwritten. Fix before publication.
7. The sit-still test (sandbox). If the party holes up and does nothing for three in-game weeks, what has happened? The answer must be "the country has changed in ways the party now has to live with," not "nothing." If the country pauses when the party pauses, it is not a sandbox.
8. The shortcut test (sandbox). Open your "When the Party Shortcuts" section (or, in a chapter, your failure-mode section). Does every shortcut have a true consequence, not a reset? Is there a line for "the party does something nobody planned for"? A shortcut answered with a reset is a railroad in disguise.

A note on iteration

The workflow is presented in order, but no draft survives first contact with this order on the first pass. Expect to:

• Draft a thin Arc I before factions and tracks are finalised, then loop back and revise the cold open once the tracks are named.
• Discover during scene drafting that an NPC needs a different motivation; revise the cast entry rather than forcing the scene.
• Find during the anti-railroad pass that an entire arc is load-bearing on a single NPC; redistribute, then re-audit.

Iteration is not failure. It is how the adventure assembles. The order in the workflow is the order of the final pass, not the only pass.

Safety and content discipline

Adventures touching difficult content (violence against specific groups, body horror, intimate threat, real historical atrocities) carry a structural responsibility to the GMs and players who will run them.

• Declare content types in the GM notes at the top of the adventure, before the cold open. Use a short bracketed list (e.g., "Contains: lethal violence, references to historical ethnic displacement, brief on-page death of a child NPC").
• Recommend session zero calibration for any adventure that contains content beyond the system's baseline assumptions.
• Reference an in-play safety tool the table can use (X-card, Lines and Veils, Script Change). Do not require a specific one. Confirm players know how to use whichever the table picks.
• For horror specifically, see references/horror-design.md — safety calibration is not optional in horror, it is part of the design contract.

What this skill is not

• Not a system. No mechanics, no stat blocks, no resolution rules. The combat-encounter block is a structural template, not a ruleset. Plug in your system's numbers.
• Not a setting. No genre commitments. Western, fantasy, sci-fi, modern, horror, post-apocalyptic — the architecture is the same.
• Not a writing-style guide. For prose voice, load a separate writing skill. This bible governs structure, branching, and design — not sentence-level craft.
• Not a tournament module template. Tournament adventures optimize for completion in a fixed time. Sandbox openers optimize for replayability and downstream consequence. This bible serves the latter.