Help us improve
Share bugs, ideas, or general feedback.
From skills-for-humanity
Diagnoses and repairs dialogue for subtext, voice differentiation, exposition, and forward momentum. Use when dialogue sounds on-the-nose or characters sound the same.
npx claudepluginhub human-avatar/skills-for-humanityHow this skill is triggered — by the user, by Claude, or both
Slash command
/skills-for-humanity:s4h-writing-dialogueThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Dialogue fails in two primary ways: it says exactly what it means (no subtext), or it explains things the characters already know (exposition dressed as conversation). Both failures produce the same effect on the reader — a sense of flatness, of the author's hand visible, of characters becoming mouthpieces rather than people.
Rewrites screenplay dialogue to improve subtext, differentiate character voices, and remove on-the-nose writing. Useful for flat or interchangeable dialogue in final drafts.
Diagnoses and repairs flat, purposeless scenes using the want/obstacle/outcome framework. Every scene must change the story's state.
角色质量检查工具,检测人设扁平化、前后不统一、配角工具人化等问题。当用户说"角色检查"、"人设检查"、"人物检查"、"检查角色"时自动激活。
Share bugs, ideas, or general feedback.
Dialogue fails in two primary ways: it says exactly what it means (no subtext), or it explains things the characters already know (exposition dressed as conversation). Both failures produce the same effect on the reader — a sense of flatness, of the author's hand visible, of characters becoming mouthpieces rather than people.
Real conversation is oblique. People in conflict rarely say "I'm angry with you." They talk about something else. The anger is present in what they say about the dishes, the schedule, the way someone looks. The subtext is the scene — the surface conversation is the vehicle for it. When dialogue is on the nose, the subtext and the surface are the same thing, which means there is no scene — only information delivery.
Voice differentiation is the second major failure point. Every character has a distinct rhythm of thought, a vocabulary range shaped by their history, evasion patterns specific to their wound, and a relationship to silence. When characters sound interchangeable, the dialogue's only differentiator is the attribution ("he said / she said") — which means the characters, on the page, don't exist yet.
Step 1: Speaker Goals — Surface vs. Subtext For each speaker in the exchange, identify:
Framing check: Confirm the specific dialogue before continuing. State what you've identified — the exchange being analysed, the speakers involved, and the apparent dramatic situation — in one sentence, then use AskUserQuestion:
These should be different. If surface goal and subtext goal are identical, the speaker is saying exactly what they mean — the subtext has collapsed into the text, and the scene has lost its tension.
Step 2: Voice Differentiation Analyse each speaker's distinct voice characteristics:
If the voices are interchangeable — if you could swap the speakers' lines without the scene changing — the characters have not yet been individualised.
Step 3: Exposition Check Flag any lines where characters explain things they both already know, solely for the reader's benefit. This is the "As you know, Bob" problem: "As you know, we've been married for fifteen years and your father never approved of me." Neither character needs to be told this. It's only there for the reader — and the reader knows it.
Exposition that is genuinely needed can be delivered through conflict (characters who have different versions of the same event), curiosity (a character who genuinely doesn't know something), or revelation (information emerging under pressure, not volunteered).
Step 4: Scene Function What is the dialogue doing? Mark all that apply:
If the dialogue serves no discernible function — if it is simply exchange — it should be cut or redirected.
Step 5: Forward Momentum Does the dialogue advance the scene, or does it stall it? Diagnose the rate at which the scene's situation changes through the dialogue. If the first and last lines of the dialogue exchange are in the same state as each other, the dialogue is stalling.
Before proceeding, use the AskUserQuestion tool. State your interpretation of the situation in 1–2 sentences — what is being analyzed and what the core question is — then ask:
Proceed based on their selection. If the user reframes, incorporate the correction before running any analysis.
Speaker Goals:
Voice Differentiation:
Exposition Flags: [Quoted lines + why they're exposition + how to deliver the information legitimately]
Scene Function: [Conflict / Revelation / Bonding / Negotiation / Deflection — mark all present / FLAG if none]
Forward Momentum: [Does the scene advance through dialogue? Where does it stall?]
Line-by-Line Notes: [Specific notes on strongest exchanges and weakest — quote and diagnose]
Rewrites for Flagged Lines: [Specific rewrites that restore subtext, differentiate voice, or remove exposition]
/s4h-writing-scene-construction for diagnosing the dialogue within its scene context — dialogue problems are often scene-construction problems./s4h-writing-character-development because voice flows from character: the wound and defence determine how a character speaks, evades, and what they're incapable of saying directly./s4h-writing-voice-consistency when the issue is that a character's voice is inconsistent across the manuscript rather than within a single scene.After delivering this output, use AskUserQuestion to offer the next move:
/s4h-writing-voice-consistency — Ensure dialogue is voice-consistent throughout/s4h-writing-character-development — Use dialogue to deepen character/s4h-writing-scene-construction — Embed the dialogue in its scene