audit-narrative-state

Audit Narrative State

Treat narrative implementation as a state machine and data-flow system.

This audit focuses on explicit or inferable state such as:

• quest phase
• relationship state
• alive/dead/present status
• faction allegiance
• knowledge flags
• choice memory
• inventory/story objects
• world conditions
• romance state
• reputation
• story chapter

Core Checks

Initialization

• state read before it can be set
• ambiguous defaults
• old-save migration assumptions

Transition validity

• impossible jumps
• missing transition path
• state can regress unintentionally
• terminal state can reopen unexpectedly

Contradictory combinations

Examples:

• dead = true and active_companion = true
• quest = completed and objective still active

Do not assume combinations are invalid merely because they look strange; look for authored interpretation.

Memory

• meaningful choice is set but never read
• state is read in one branch and ignored after reconvergence
• consequence resets without narrative explanation

Reachability

• unreachable content
• conditions that can never all be true
• state that can be entered but never exited

Data-flow

• set-but-never-read
• read-but-never-set
• duplicate flags representing the same concept
• two variables that can disagree about the same truth

Combinatorial risk

Persistent independent variables may multiply downstream variants even when the visible branch graph is small.

Procedure

1. Inventory variables/states.
2. Identify possible values and defaults.
3. Locate sets, reads, and resets.
4. Model valid transitions.
5. Enumerate or sample combinations where useful.
6. Trace critical player choices through reconvergence.
7. Identify unreachable, contradictory, or unhandled states.
8. Rank by player impact and production/QA risk.

Output

State inventory

State	Values	Default	Set at	Read at	Reset at

Transition issues

Contradictory / unhandled combinations

Dead or unused state

Memory failures

Reachability issues

Complexity hotspots

For each finding include:

• evidence
• severity
• confidence
• affected content
• smallest repair category

Repair categories:

• initialize explicitly
• merge duplicate state
• add/remove transition
• gate content
• preserve state across foldback
• add handler for combination
• remove dead flag
• convert multiple booleans to enum/state machine

Severity

Critical for progression blockers, impossible saves, or major narrative contradictions. High for broken consequence/quest logic. Medium for latent combinations or wasted content. Low for cleanup and maintainability risks.

Final Quality Check

Separate logical validity from authorial desirability. A bitter romance can coexist with hostility if the narrative supports it; an unhandled combination is the problem, not complexity itself.