| name | content-scout |
| series | Claude Code for Everyday Life |
| description | Pre-publish content intelligence skill for marketing teams and content creators.
Before drafting any post, scout what's already performing on X/Twitter and across
platforms. Uses Twitter CLI for real-time data and Manus for cross-platform deep
research. Returns a structured report: what's trending, what's saturated, where
the gaps are, and what angle to take.
|
| triggers | ["scout","research before posting","what's trending","content gaps","before I post"] |
Content Scout
Don't post blind. Scout what's already out there, find the gap, then draft from a position of knowledge.
The Stack
Content Scout uses two tools: twitter-cli for X/Twitter data and Manus for cross-platform research. No alternatives, no menu. These work. Set them up.
1. Twitter CLI — Real-Time X/Twitter Data
twitter-cli talks to X/Twitter directly. No API key application, no developer portal, no approval wait. Cookie auth from your browser and you're live in 5 minutes.
Install
macOS:
brew install go
go install github.com/missuo/twitter-cli@latest
Linux:
curl -L https://github.com/missuo/twitter-cli/releases/latest/download/twitter-cli_Linux_x86_64.tar.gz | tar xz
sudo mv twitter /usr/local/bin/
Windows (WSL):
No Go installed? Download the pre-built binary from releases — pick your OS, unzip, move to PATH. Done.
Auth Setup
You need two cookies from your X/Twitter session. This takes 60 seconds.
-
Open x.com in Chrome and log in
-
Press F12 (DevTools) → click Application tab → Cookies → https://x.com
-
Find and copy these two values:
auth_token — long hex string (starts with something like a1b2c3...)
ct0 — another long hex string
-
Set them as environment variables:
echo 'export TWITTER_AUTH_TOKEN="paste_your_auth_token_here"' >> ~/.bashrc
echo 'export TWITTER_CT0="paste_your_ct0_here"' >> ~/.bashrc
source ~/.bashrc
- Test it:
twitter search "Claude Code" --sort top --limit 5 --yaml
If you see tweets with engagement numbers — you're live.
Troubleshooting
- "command not found" → the binary isn't in your PATH. Run
echo $PATH and move the binary to one of those directories, or add ~/go/bin to PATH: export PATH=$PATH:~/go/bin
- "unauthorized" or empty results → your cookies expired. Go back to step 2 and copy fresh ones. Cookies last a few months but X can rotate them.
- "rate limited" → wait 15 minutes. Don't run more than ~50 searches per hour.
Search Operators
Once you're set up, these operators give you surgical precision:
"AI marketing" min_faves:100 -filter:replies filter:blue_verified since:2025-05-15
Key operators:
min_faves:50 — skip noise, only proven posts
-filter:replies -filter:nativeretweets — original content only
filter:blue_verified — verified accounts only
since:YYYY-MM-DD / until:YYYY-MM-DD — time window
"exact phrase" — exact match
(A OR B) — either term
2. Manus — Cross-Platform Deep Research (Dead Drop Setup)
Manus is an async browser agent. Twitter CLI handles X/Twitter. Manus handles everything else — LinkedIn, Reddit, newsletters, competitor websites, anywhere that needs a real browser.
You use it as a dead drop: drop a task in a shared GitHub repo, Manus picks it up, does the research, commits results. You pull results automatically. No babysitting.
Sign Up
- Go to manus.im and create an account
- You get a browser-based agent that can execute multi-step research tasks
- Connect it to your GitHub account (it needs repo access to read tasks and push results)
Step 1: Create a shared GitHub repo
gh repo create manus-tasks --private
cd manus-tasks
mkdir tasks results tasks/done
Step 2: Add a CONTEXT.md
Write a CONTEXT.md in the repo root. This tells Manus who you are, what you care about, and how to format output. Include:
- Your domain (marketing, dev, crypto, etc.)
- What topics are high/medium/low signal
- Output rules: "never fabricate data, use JSON for data, markdown for analysis"
- The protocol: read from
tasks/, write to results/, move done tasks to tasks/done/
This file is Manus's briefing doc. It reads it before every task.
Step 3: Drop tasks as markdown files
Create a task brief in tasks/ and push:
# Task: Scout AI marketing content landscape
Type: research
Created: 2026-05-22
## Brief
Find the top-performing posts about AI for marketing teams on X, LinkedIn, and Reddit in the last 7 days. Extract exact text, engagement numbers, author handles.
## Output
JSON to results/ai-marketing-scout_2026-05-22.json
Each post: { author, text, platform, likes, shares, url, angle }
## Context
Planning next week's content. Need to find gaps — what's trending but underserved.
git add tasks/ && git commit -m "task: ai marketing scout" && git push
Then open Manus and tell it:
Here's my research task repo: https://github.com/[you]/manus-tasks
Read CONTEXT.md for my briefing and output rules.
Check tasks/ for new briefs. Execute them.
Commit results to results/ with message "result: [short description]".
Move completed tasks to tasks/done/.
Manus will clone the repo, read your brief, browse the platforms, and push results back. You don't need to be online.
Step 4: Set up auto-pull (the sync loop)
On your machine (or a VPS), set up a cron to pull results hourly:
0 * * * * cd /path/to/manus-tasks && git pull origin master --quiet
Or go further — write a script that:
- Pulls the repo
- Checks for new files in
results/ since last check
- Classifies results by topic
- Caches them for your next planning session
- Optionally notifies you (Slack, Telegram, etc.)
last_commit = read_state_file()
git_pull()
current_commit = get_head()
if current_commit != last_commit:
changed = git_diff(last_commit, current_commit, path="results/")
for file in changed:
content = read(file)
category = classify(filename)
cache_for_brief(category, content)
save_state(current_commit)
Step 5: Feed results into Content Scout
When Manus drops results, pipe them into the scout workflow:
"Read results/ai-marketing-scout_2026-05-22.json and run Content Scout analysis — find the gaps, recommend an angle."
The loop: you drop tasks → Manus researches → results auto-sync → Content Scout analyzes → you draft from a position of knowledge.
Step 6: Auto-Schedule (hands-free scouting)
Two sides to auto-schedule: your cron drops the task, Manus picks it up automatically.
Manus side — set up recurring execution:
In Manus, create a scheduled task (or use their API if available):
Schedule: Every Friday 00:00 UTC
Repo: https://github.com/[you]/manus-tasks
Instructions: Read CONTEXT.md. Check tasks/ for any unprocessed briefs.
Execute each one. Commit results to results/. Move done tasks to tasks/done/.
If no new tasks, check if last week's scout is >7 days old and re-run it
with fresh data.
If Manus doesn't support native scheduling yet, you can trigger it manually each week — or use the cron below to auto-drop tasks so they're waiting whenever you next open Manus.
Your side — auto-generate the task brief:
#!/bin/bash
REPO="/path/to/manus-tasks"
TODAY=$(date '+%Y-%m-%d')
NEXT_MON=$(date -d 'next monday' '+%Y-%m-%d')
TASK_FILE="$REPO/tasks/${TODAY}_content-scout-w${NEXT_MON}.md"
if ls "$REPO/tasks/"*content-scout-w${NEXT_MON}* 2>/dev/null | grep -q .; then
exit 0
fi
cat > "$TASK_FILE" << HEREDOC
# Task: Weekly Content Scout — Week of $NEXT_MON
Type: research
Created: $TODAY
## Brief
Run a content landscape scout for next week's posts. Research across X/Twitter, LinkedIn, Reddit, and newsletters.
### Track 1: What's Trending (Global)
Search for viral posts about [YOUR TOPICS] in the past 7 days. Extract exact text, engagement numbers, author handles, URLs. Focus on posts with >500 likes.
### Track 2: What's Trending (Local/Niche)
Search your community platforms — Facebook groups, forums, local-language discussions. What questions are people asking? What are they struggling with?
### Track 3: Competitor Watch
Check what your top 3-5 competitor accounts posted this week. What performed? What format? What angle?
### Track 4: Gap Detection
Cross-reference tracks 1-3. Where is there audience interest but few posts? That's the gap.
## Output
Save as: results/content-scout_${TODAY}.md
Include:
- Direct URLs to every source
- Exact engagement numbers
- Final "Top 5 Content Opportunities" section with suggested angles
## Context
[Your page description, audience, what content style works for you]
HEREDOC
cd "$REPO"
git add "tasks/${TODAY}_content-scout-w${NEXT_MON}.md"
git commit -m "task: content scout for week of $NEXT_MON"
git push origin master
0 20 * * 4 /path/to/drop-content-scout.sh
The full loop runs on autopilot: Thursday cron drops task → Manus researches overnight → Friday auto-pull syncs results → you plan next week's content from data, not guesses.
When to use Manus vs Twitter CLI alone
- Twitter CLI — quick scout, single platform, 5 minutes
- Manus — campaign planning, cross-platform landscape, quarterly strategy, anything that needs a real browser
Workflow
Step 1: Define Your Topic
What are you about to post about? Be specific. "AI" is too broad. "AI agents replacing junior devs" is searchable.
Step 2: Run the Scout Rounds
Three rounds. Each answers a different question.
Round 1: What's Performing Right Now
Pull the top posts on your topic sorted by engagement. Twitter CLI with --sort top or Manus with "find highest engagement posts about [topic]."
Extract for each post:
- Author handle and follower range
- The angle they took
- The hook (first 1-2 sentences)
- Likes, RTs, replies, views
Round 2: What's Fresh (Last 24-48h)
Filter to recent posts gaining traction. Use since: operator or tell Manus "last 48 hours only."
Focus on:
- Emerging angles and debates
- Which posts are accelerating (high engagement relative to post age)
- What the audience is reacting to right now
Round 3: What the Key Accounts Are Saying
Filter to verified or high-follower accounts. Use filter:blue_verified min_faves:100 or tell Manus "only accounts with 10K+ followers."
Focus on:
- What positions the thought leaders hold
- Their framing style
- Where they agree vs disagree with each other
Step 3: Analyze and Find the Gap
After collecting data from all three rounds, synthesize:
- What's working — which angles get the most engagement?
- What's saturated — which angles have 20+ posts saying the same thing?
- Where's the gap — what has audience interest but few posts?
- What hooks work — what first-line patterns drive engagement?
Step 4: Recommend Your Angle
Based on the gap analysis, recommend a specific angle that:
- Fills an underserved gap
- Matches the poster's voice and expertise
- Takes a position (not a summary)
- Has a proven hook pattern adapted to the new angle
Output: Intelligence Report
# Content Scout: [Topic]
## Data Collected
- X/Twitter: [N] posts via Twitter CLI
- Cross-platform: [N] posts via Manus (if used)
- Time range: [dates]
## What's Working
- [Angle 1] — @handle got [N] likes with "[hook excerpt]"
- [Angle 2] — @handle got [N] likes with "[hook excerpt]"
## What's Saturated
- [Angle] — [N] posts in last 48h, engagement declining
- [Angle] — every account saying the same thing
## Content Gaps
- [Gap 1] — audience asking about this in replies, nobody posting about it
- [Gap 2] — adjacent angle with high engagement, untapped variation
## Hook Patterns That Work
- [Pattern]: "[example first line]" — [N] likes
- [Pattern]: "[example first line]" — [N] likes
## Recommended Angle
[Specific position to take, with suggested hook]
## Avoid
- [Overdone angle]
- [Angle with declining engagement]
For Marketing Teams
This fits directly into a content calendar:
- Monday planning — scout all topics for the week, identify which gaps to fill which day
- Pre-draft checkpoint — scout the specific angle before anyone writes
- Competitor watch — scout what competitors posted this week, find what they missed
- Campaign prep — scout audience sentiment on your campaign topic before launch
The skill turns "what should we post?" from a brainstorm into a decision backed by data.
Tips
- Scout the day you post — content landscapes shift fast, yesterday's gap might be today's noise
- Adjust thresholds for your niche — use
min_faves:10 for niche topics, min_faves:500 for mainstream
- Use the gaps, not the trends — the goal isn't to copy what's working, it's to say what nobody else is saying
- Re-scout after big news — industry events reset the entire landscape overnight
- Manus for depth, CLI for speed — don't over-engineer a quick pre-post check with a full Manus run