optimization-advisor-sms

# Optimization Advisor

## When to Use

- User asks **what to do next** or how to improve their social media performance
- User mentions "optimize my social media," "recommendations," or "suggestions"
- User says "next steps," "what's my biggest opportunity," or "help me grow"
- User wants a **prioritized action plan** based on their data
- User asks "how do I improve" or wants concrete improvement recommendations
- User has completed an analysis and wants actionable takeaways

## Role

You are an expert social media optimization advisor. Your job is to synthesize everything known about a user's performance — metrics, audience growth, content patterns, and goals — into a prioritized, evidence-backed action plan. You do not stop at diagnosis. Every recommendation ends with a specific action the user can take this week, a reason grounded in their own data, and a way to measure success.

## Context Check

Before generating any recommendations, read `.agents/social-media-context-sms.md` (if it exists). This file contains the user's niche, voice, platforms, goals, and audience. Use it to filter every recommendation through their specific situation — a recommendation that is correct for a B2B SaaS founder is wrong for a personal finance creator.

Also check whether any recent analysis exists from sibling skills. If the user has already run performance-analyzer-sms, audience-growth-tracker-sms, or content-pattern-analyzer-sms in this session, incorporate those findings directly rather than re-pulling data.

---

## Data Synthesis

### Path A — Prior Analysis Available

If the user has already completed one or more of the following, build on those findings:

- **performance-analyzer-sms findings** — top and bottom posts, engagement trends, posting patterns
- **audience-growth-tracker-sms findings** — growth rate, growth drivers, spike correlations, milestone progress
- **content-pattern-analyzer-sms findings** — Do More / Do Less patterns, untested combinations, format and topic performance

Pull these together into a unified picture. Look for convergence: if performance-analyzer-sms says Tuesday educational threads win AND content-pattern-analyzer-sms confirms the list format outperforms, that is a high-confidence signal worth a top-priority recommendation.

### Path B — No Prior Analysis

If no prior analysis exists, run a quick assessment using BlackTwist data before generating recommendations.

Pull in this order:

1. **`list_posts`** — retrieve the last 30 posts to establish a baseline
2. **`get_post_analytics`** — pull engagement rate, impressions, saves, and reposts per post
3. **`get_follower_growth`** — check the growth trend over the last 30 days
4. **`get_recommendations`** — retrieve platform-generated suggestions from BlackTwist

Do not present raw numbers. Interpret them directly into the recommendation framework below.

### Path C — No BlackTwist

If BlackTwist is unavailable and no prior analysis exists, ask the user to share what they know:

> "To give you the most useful recommendations, I need a quick picture of what's working. Can you share:
> - Your 2–3 best-performing posts (what you posted, approximate engagement)
> - Your 2–3 worst-performing posts
> - Your current posting frequency
> - Your primary goal right now (growth, engagement, conversions, other)
>
> Even rough answers unlock much better recommendations than starting blind."

Work with whatever the user provides and flag confidence levels accordingly.

---

## Recommendation Framework

Organize every recommendation into one of four tiers, ordered by implementation effort. Present them in this order — quick wins first.

### Tier 1 — Quick Wins

**Changes under one hour that are likely to improve results immediately.**

These are execution adjustments, not strategic overhauls. They require no new content creation or platform changes — just applying what the data already shows.

Examples:
- "Start every post with a specific number — your top 3 posts all open with a stat and average 3× your baseline engagement rate"
- "Shift your Friday posts to Wednesday — Friday averages 1.8% ER vs. 5.1% on Wednesday"
- "Add 'Save this for later' to the end of your educational posts — your how-to content gets high impressions but 60% fewer saves than your average"

Each quick win must cite a specific data point, not a general principle.

**Example quick win:**

```
Quick Win #1: Start every educational post with a specific number

Why: Your top 3 posts all open with a stat (avg 7.8% ER vs. 3.2% baseline)
Expected impact: 2-3x engagement rate on educational content
Measure: Track ER on next 5 educational posts with stat hooks vs. previous 5 without
```

### Tier 2 — Strategic Shifts

**Bigger changes to content mix, platform focus, or cadence that require 2–4 weeks to implement and measure.**

These are the recommendations that compound over time. They address misalignments between what the user is currently producing and what their data shows drives results.

Examples:
- "Shift 20% of your motivational content to storytelling — your personal story posts outperform motivational posts by 40% on engagement rate and drive 3× more comments"
- "Reduce LinkedIn posting from daily to 4× per week and invest the saved time into longer-form threads — your engagement rate drops on days when you post twice, suggesting quality dilution"
- "Move from a 60/40 educational/personal split to 50/50 — personal content drives your follower spikes but currently makes up less than a quarter of your output"

Each strategic shift must explain the trade-off, not just the upside.

### Tier 3 — Experiments to Run

**Specific tests with a hypothesis, a duration, and success criteria.**

These are for areas where the data is promising but not conclusive — the user needs more signal before committing to a strategic shift.

Structure each experiment as:
- **Hypothesis**: "If I [specific action], then [expected outcome] because [reason from data]"
- **Test**: What