---
name: "reply_game_playbook"
description: "The substantive-reply playbook: how to pick threads worth your time and write replies that add real signal — never 'great post'. Use when drafting replies on X, forums, comment threads, or community posts; when building a reply strategy for an account; or when reviewing reply drafts for quality."
metadata: { "includeInPrompt": true }
---

# Reply Game Playbook

## Purpose

Turn replying into a repeatable craft: **pick the right threads → add real signal → keep
the conversation going**. Works on X, forum threads, blog comments, Discord/Reddit threads —
anywhere a thread of strangers exists. For X-specific algorithm mechanics and account
growth strategy, see the `x_growth_playbook` skill instead; this one is the craft of the
reply itself.

## Workflow

### 1. Select the thread (read `references/thread-selection.md`)

Reply where the audience already gathers, not where you wish they would. The short form:

- **Audience fit first.** The author's audience should overlap your target readers.
  2–100× your own audience size is the sweet spot; celebrities add noise, ghost towns add nothing.
- **Fresh and moving.** Target threads under ~2 hours old that are still gaining replies.
  Practitioner consensus (not algorithm fact): the first 30–60 minutes give the best visibility.
- **Skip the traps.** Rage-bait threads, quote-tweet dogpiles, and threads you don't
  genuinely understand well enough to add to. A reply that teaches nothing teaches about you.

### 2. Read before you write

Read the **entire thread**, not just the top post. Note: what the author actually claimed
(not what you wish they'd claimed), what the top replies already said (never repeat them),
and what question the thread is circling without asking. The gap is your opening.

### 3. Draft the reply (read `references/reply-archetypes.md`)

Every reply follows **validate → add → invite**:

1. **Validate** in ≤10 words: show you read it ("You're right that dormancy matters —").
2. **Add** one unit of signal: a number, a first-hand observation, a sharper formulation,
   a concrete example, or a respectful counterpoint. This is the whole reply's reason to exist.
3. **Invite** (optional): end with the honest question the thread is dodging, or a clear
   opening for the author to respond. Conversation, not applause.

Keep it short enough to read in under 15 seconds. If it can't stand alone as its own
post, it's not good enough to be a reply.

### 4. Match the register

Serious thread → serious reply. Joke thread → wit, not a lecture. Technical thread →
precise language, no fluff. The author's voice sets the bar: write one notch crisper,
never one notch louder. Never match hostility — sharp is fine, mean is a liability.

### 5. After you post: stay in the conversation

- Answer substantive replies to your reply with substance. The back-and-forth is where
  trust forms — and on X, author engagement on a reply is the strongest visibility signal
  in the open-sourced ranking model.
- Never argue for sport. If the exchange stops being informative, exit politely.
- Log what worked: which archetypes and which threads earned real conversation. Repeat those.

## Output Contract

When asked to **draft replies**, deliver for each thread:
1. **The reply** — final text, within platform character limits (280 chars on X unless
   the account is verified).
2. **Rationale (1–2 lines)** — which archetype it uses and what signal it adds.
3. **Risk note** — if the reply could read as spam, flattery, or off-register, say so.

When asked to **review replies**, deliver: verdict per reply (ship / rewrite / drop) with
one line each on what to cut or sharpen.

## Operating Rules

1. **Never "great post".** Never "this!!", "100%", "so true", bare emoji, or restating the
   author's point in different words. Zero-signal replies are invisible at best and
   reputation damage at worst.
2. **Standalone quality test.** Before posting, ask: "Would this work as its own post?"
   If no, rewrite or drop it.
3. **One signal per reply.** Don't bundle three insights — pick the strongest.
4. **Credit sources.** If the insight isn't yours, say whose it is. Plagiarism of a reply
   is still plagiarism.
5. **No farming patterns.** One good reply per thread; never spray multiple replies across
   an author's timeline, never copy-paste one take across accounts. Spam patterns get
   suppressed and earn mutes — the negative signal that kills distribution fastest.
6. **Honesty about provenance.** Verified algorithm facts (X's open-sourced ranking model)
   ≠ practitioner opinion (timing windows, archetype effectiveness). Never present
   opinion as fact, never invent multipliers or engagement statistics.
7. **This skill drafts; it never auto-posts.** A human or an explicitly authorized
   posting flow makes the final call.