Hooks

Skill Epistates/sparX/.claude/skills/hooks

Phoenix-powered X/Twitter content engine for Claude Code: draft, optimize, score, post, engage, and analyze using real 2026 algorithm weights. No APIs needed.

Install

npx -y skills add Epistates/sparX --skill hooks

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate scroll-stopping opening lines for X posts. Use when the user needs hook ideas, opening lines, first lines, or attention-grabbing openers for tweets.

SKILL.md

3.3 KB, as published. Nobody here has run it

Generate Scroll-Stopping Hooks

The first 8-12 words of an X post determine whether anyone reads the rest. Generate multiple high-impact hook options optimized for the Phoenix algorithm's dwell time signal.

Input

The user provides:

A topic they're posting about
A rough idea of the post content
An existing hook they want alternatives for
A URL — read the content via WebFetch to understand the subject, then generate hooks based on the actual material (see ../../../reference/url-reading.md)

Process

Step 1 — Understand the Content

What is the core message? Who is the audience? What emotion should the hook trigger?

Step 2 — Read Hook Patterns

Read patterns.md for the full hook pattern library.

Step 3 — Generate 7-10 Hooks

Generate hooks across multiple pattern categories:

Curiosity Gap — Creates an information gap the reader must fill
Specific Number — Leads with a concrete, impressive metric
Contrarian — Challenges conventional wisdom
Pain Point — Names a frustration the audience feels
Before/After — Shows transformation in minimal words
Social Proof — Leads with credibility markers
Question — Poses something the reader can't ignore

For each hook:

Keep to 8-12 words (the visible preview before truncation)
Test: Would this stop YOUR scroll?
Ensure it's honest — don't promise what the post doesn't deliver

Step 4 — Rank and Annotate

Rank all hooks by predicted effectiveness and annotate:

Best for dwell — which hook will make people read the full post?
Best for replies — which hook triggers an urge to respond?
Best for bookmarks — which hook signals "save-worthy" content?
Safest — which hook has lowest negative signal risk?
Boldest — which hook has highest ceiling but more risk?

Step 5 — Output

Present all hooks in a scored table, ranked by Phoenix Score. The score represents what a full post would score with that hook (assuming a solid body + CTA):

#  Hook                                                Score   Best for
1. "2.3× faster than MLX — on a MacBook."              8.4/10  Dwell + bookmarks
2. "Your MacBook can fine-tune a 7B model in 12 min."  8.1/10  Specificity + curiosity
3. "Why I rewrote LLM inference in pure Rust."          7.9/10  Curiosity gap
4. "Most devs are overpaying for GPU cloud. Here's why." 7.6/10  Pain point + broad reach
5. "The local AI tipping point just arrived."           7.2/10  Boldest (high ceiling, more risk)
...

Also include:

Annotations per hook (best for dwell / replies / bookmarks / safest / boldest)
Recommended top pick with rationale
The complete first 2 lines (hook + bridge to body) for the top 3

Rules

Never use engagement bait ("You won't believe...")
No clickbait that the content doesn't deliver on
Avoid jargon in the hook unless the audience is entirely specialists
Specificity always beats vagueness ("2.3× faster" > "much faster")
The hook must be true