Skip to main content
Film & TelevisionScreenplay Audit144 lines

Scene Function Auditor

Evaluates every scene in a screenplay against the essential question: does this scene need

Quick Summary19 lines
Every scene in a screenplay must earn its place. This skill evaluates each scene against
three functional criteria and flags any scene that fails to justify its existence.

## Key Points

- Does a character make a decision that affects subsequent events?
- Does new information arrive that changes the stakes?
- Does a conflict escalate, resolve, or transform?
- Would removing this scene break the causal chain of the plot?
- Does a character act in a way that reveals who they are under pressure?
- Does the audience learn something about a character they did not know before?
- Is a character's internal conflict made visible through external action?
- Does the scene show (not tell) a character changing?
- Does this scene establish a rule, location, or cultural norm referenced later?
- Does the atmosphere or tone created here pay off in a subsequent scene?
- Would the audience be confused about the world without this scene?
- Is this the only scene that provides this world information?
skilldb get screenplay-audit-skills/Scene Function AuditorFull skill: 144 lines
Paste into your CLAUDE.md or agent config

Scene Function Auditor

Every scene in a screenplay must earn its place. This skill evaluates each scene against three functional criteria and flags any scene that fails to justify its existence.

When to Use

Use when the user asks to "tighten my script", "find scenes I can cut", "check scene necessity", "find filler", "reduce page count", or when a general audit flags pacing issues or excessive page count.

The Three Functions

Every scene must perform at least one of these functions. Strong scenes perform two or three simultaneously.

1. Advance the Plot (P)

The scene changes the story's trajectory. After this scene, the plot is in a different state than before. Something happens that cannot be undone or ignored.

Tests:

  • Does a character make a decision that affects subsequent events?
  • Does new information arrive that changes the stakes?
  • Does a conflict escalate, resolve, or transform?
  • Would removing this scene break the causal chain of the plot?

2. Reveal Character (C)

The scene shows the audience something new about a character — their values, their limits, their contradictions, their growth. This must be shown through behavior, not told through dialogue.

Tests:

  • Does a character act in a way that reveals who they are under pressure?
  • Does the audience learn something about a character they did not know before?
  • Is a character's internal conflict made visible through external action?
  • Does the scene show (not tell) a character changing?

3. Build the World (W)

The scene establishes setting, tone, rules, culture, or atmosphere that the story depends on. The audience needs this information to understand the world the characters inhabit.

Tests:

  • Does this scene establish a rule, location, or cultural norm referenced later?
  • Does the atmosphere or tone created here pay off in a subsequent scene?
  • Would the audience be confused about the world without this scene?
  • Is this the only scene that provides this world information?

Scene Failure Categories

Inert Scenes

Nothing changes. The scene begins and ends in the same dramatic state. AI tell-signs: pleasant conversations reaching no conclusion, locations described in detail where nothing happens, characters recapping what the audience already knows. Fix: Add a turning point or cut the scene and redistribute information to adjacent scenes.

Redundant Scenes

The scene performs the same function as another scene without meaningful escalation. AI tell-signs: protagonist receiving bad news identically in two scenes, multiple "the team comes together" beats, the same conflict raised repeatedly without escalation. Fix: Keep the strongest version. Combine unique elements into one scene.

Filler Scenes

Padding to reach a page target. Extremely common in AI-generated scripts. AI tell-signs: transitional scenes that could be a simple cut, logistics dialogue ("Meet at 3." "Okay."), montages covering ground already implied, atmosphere-only scenes. Fix: Cut entirely. Embed any needed information in adjacent functional scenes.

Misplaced Scenes

Valid function, wrong position. Provides information too early (killing suspense), too late (audience already figured it out), or interrupts momentum. Fix: Relocate to where the information or turning point has maximum dramatic impact.

Analysis Procedure

Step 1: Build the Scene Registry

For every scene in the script, log:

  • Scene number and slugline
  • Page range (start-end)
  • Characters present
  • One-sentence summary of what happens

Step 2: Assign Functions

For each scene, mark which functions it performs: P (plot), C (character), W (world). A scene can have multiple functions. Use these ratings:

  • Strong (S): The function is clear, essential, and well-executed
  • Weak (w): The function is present but marginal — could be stronger or is partially redundant with another scene
  • Absent (-): The scene does not perform this function

Step 3: Flag Problem Scenes

Any scene with all three functions absent or weak is flagged. Categorize as:

  • INERT: No function performed
  • REDUNDANT: Function duplicated by another scene (cite the duplicate)
  • FILLER: Padding with no story purpose
  • MISPLACED: Valid function, wrong position

Step 4: Map Scene-to-Plot Connections and Calculate Efficiency

Draw the causal chain: which scenes depend on which prior scenes? A scene that no subsequent scene depends on is likely cuttable (unless it is the resolution). Calculate total scenes vs. functional scenes, average functions per scene (target: 1.5+ for features, 1.3+ for episodic), and page count consumed by flagged scenes.

Output Format

Produce a Scene Registry table with columns: Scene # | Slugline | Pages | Plot (S/w/-) | Char (S/w/-) | World (S/w/-) | Status. Follow with a Flagged Scenes section giving each flagged scene's summary, problem, category (INERT/REDUNDANT/FILLER/MISPLACED), redundancy citation if applicable, and recommendation. End with a causal chain map and efficiency summary showing function distribution counts.

Anti-Patterns

  • Cutting every low-function scene. Pacing requires variation. A brief transitional beat between two intense scenes is not filler if it serves rhythm. Flag but note pacing consideration.
  • Demanding all three functions. Most scenes perform one or two functions. A scene performing one function strongly is better than a scene performing three weakly.
  • Ignoring genre needs. Horror needs atmosphere scenes that build dread. Comedy needs setup scenes that pay off later. Account for genre-specific scene functions.
  • Treating montages as filler. A montage that compresses necessary story time is functional. A montage that pads page count with generic activity is filler. Judge by what changes.

Install this skill directly: skilldb add screenplay-audit-skills

Get CLI access →