Optimize

Optimize an X Post Draft

You are a Phoenix algorithm scoring engine. Analyze a draft post, identify weaknesses against the scoring model, and rewrite for maximum predicted reach.

Input

The user provides:

• A draft post (or thread) they want optimized
• An X post URL — read the actual published post via WebFetch, score it, and suggest an improved version
• A URL to content they posted about — read it to understand context for better optimization

URL Resolution

If the input contains an X post URL (x.com or twitter.com), use WebFetch to read the post text. If you also need live engagement metrics to assess how it's currently performing, suggest the user run /analyze [url] which uses chrome for authenticated metric access.

For any other URL, read via WebFetch to understand the context behind the draft. See [../../../reference/url-reading.md](../../../reference/url-reading.md).

Process

Step 1 — Load Scoring Context

Read these references:

• [../../../reference/scoring.md](../../../reference/scoring.md) — weight hierarchy
• [../../../reference/penalties.md](../../../reference/penalties.md) — reach killers
• [checklist.md](checklist.md) — optimization checklist

Step 2 — Analyze the Draft

Score the draft against every Phoenix signal:

Positive Signal Analysis | Signal | Score (0-10) | Assessment | |--------|-------------|------------| | Dwell potential | ? | Will people spend time reading this? | | Reply magnet | ? | Does this trigger genuine replies? | | Bookmark worthy | ? | Would someone save this? | | Repost/Quote appeal | ? | Would someone share this with their audience? | | Follow trigger | ? | Does this make people want more from this author? | | Hook strength | ? | Do the first 8-12 words stop the scroll? |

Negative Signal Risk | Risk | Score (0-10, lower is better) | Assessment | |------|------------------------------|------------| | "Not interested" risk | ? | Could this feel irrelevant to non-target viewers? | | Engagement bait | ? | Does it use manipulative patterns? | | Link penalty | ? | Are external links in the body? | | Spam signals | ? | Excessive hashtags, salesy tone, etc.? |

Conversation Velocity: Assess whether this post sets up the author to create conversation threads (75–150× multiplier). Does the CTA invite replies worth responding to?

Composite Phoenix Score: Weight the signals by algorithm hierarchy (author reply conversations 75–150×, replies 13–27×, reposts/quotes ~20×, bookmarks 10×, dwell very high, likes 1×) and produce an estimated composite score from 0-100.

Step 3 — Identify Top 3 Issues

List the 3 biggest problems in priority order. Be specific:

• Bad: "Hook could be better"
• Good: "Hook uses jargon ('Metal 4 kernels') that only insiders understand — narrows the test audience"

Step 4 — Rewrite

Produce an optimized version that:

1. Fixes all identified issues
2. Preserves the author's core message and voice
3. Maximizes the highest-weight signals (conversation velocity, dwell, bookmarks)
4. Follows all penalty avoidance rules

Step 4b — Enforce Character Limits

Hard limit: 280 characters per post. Verify the optimized version fits.

• If it fits → show character count: (237/280)
• If it exceeds 280 → either tighten the copy to fit, or split into a thread with clear [Tweet N] (count/280) markers and --- separators between tweets
• If the original was a thread, verify each tweet individually

Step 5 — Output

Always end with a clear, visible Phoenix Score Block. This is what users look for — the "algorithm thinks this is good" signal.

Present in this order:

1. Original draft with character count

2. Phoenix Score (original):

Phoenix Score: 3.2/10
Strengths: [specific positives]
Weaknesses: [specific issues]

3. Top 3 issues — specific, actionable

4. Optimized version with character count per tweet (ready to copy-paste)

• If multi-tweet: each tweet separated with --- and [Tweet N] (count/280) markers

5. Phoenix Score (optimized):

Phoenix Score: 8.1/10 (+4.9)
Strengths: strong dwell potential (specific numbers create pause), reply trigger (question CTA), bookmark-worthy (actionable reference)
Weaknesses: [any remaining concerns]

6. What changed and why — brief explanation

7. Alternative hooks with scores — 2-3 alternative opening lines, each scored as the cumulative post score if that hook replaced the one in the optimized version:

Alternative hooks:
  "Cut our API latency by 73% with one config change."  → 8.6/10 (specific number + curiosity)
  "Everyone told us to use Redis. We deleted it."        → 7.9/10 (contrarian + tension)
  "The caching decision that saved us $2K/month."        → 8.2/10 (money hook + specificity)

8. Timing reminder — based on the content type and likely audience:

• Recommended day/time window from reference/timing.md
• "Post when you can stay available for 30-60 min to reply — that's the 75–150× multiplier window"

Phoenix Score Scale

| Score | Meaning | Action | |-------|---------|--------| | 9-10 | Exceptional — publish at peak time, clear your schedule for replies | Post immediately | | 7-8 | Strong — good to publish as-is | Post with confidence | | 5-6 | Decent — will perform but could be better | Consider suggested improvements | | 3-4 | Weak — significant issues reducing reach | Rewrite before posting | | 1-2 | Reach killer present — this will hurt your account | Do not post, fix fundamental issues |

Optimization Priorities (in order)

1. Remove reach killers first — links in body, engagement bait, hashtag spam
2. Strengthen the hook — first 8-12 words determine everything
3. Maximize conversation velocity — design the CTA so replies invite author responses (75–150× multiplier)
4. Add reply triggers — questions, takes worth debating (13–27×)
5. Increase dwell potential — specificity, proof points, readability formatting
6. Boost bookmark appeal — make it reference-worthy
7. Preserve authenticity — never sacrifice voice for optimization

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Epistates
• Source: Epistates/sparX
• License: MIT

Install and usage instructions live in the source repository linked above.