Social listening with Juicer

Goal: a digest of what real people are saying about a topic right now, grounded in posts the user can click. Good for research, content ideas and articles that are not generic AI text.

Set up the search

1. Make sure Juicer is connected (see the juicer skill).
2. Ask for the topic and turn it into search terms per network:
   • Reddit: 1 to 3 keyword phrases with term_type: mentions, plus the 1 to 3 subreddits where the topic lives with term_type: channel (e.g. webdev, marketing).
   • TikTok: the keyword phrase with term_type: mentions, and the main hashtag.
   • YouTube, Instagram, X, Bluesky, LinkedIn, Pinterest: the main hashtags (#veganprotein). These networks do not support keyword search, so hashtags are the way in.
3. Show the user the plan and the credit cost before running: Reddit, X, YouTube, Bluesky and Pinterest 1 credit per call; Instagram, TikTok and LinkedIn 3; Facebook 6. A typical run (3 Reddit + TikTok + 2 hashtags on Instagram and YouTube) costs about 10 credits. Check the balance with get_account.

Run

Call get_data_posts once per term and network. Fetch more pages (cursor) only for the networks where the first page was mostly on topic.

• Reddit returning fetch_timeout ("Reddit request timed out"): Reddit is slow at times. Retry it once at the end of the run (1 credit). If it fails again, say "Reddit could not be reached, so Reddit is not covered" and never report it as "no mentions on Reddit".

Analyse

1. Drop posts that only share a word with the topic (a "#protein" post about haircare is noise). Count what you dropped.
2. Drop exact duplicates and cross-posts.
3. Group what is left into 3 to 6 themes. For each theme: one sentence, how many posts, and 1 or 2 short real quotes with links.
4. Pull out separately:
   • Questions people ask (content and FAQ ideas).
   • Complaints and frustrations (product and positioning ideas).
   • Top posts by engagement (like_count, comment_count) per network.
5. Quote only text that is in the posts. Never paraphrase a post and present it as a quote.

Digest format

Template (fill every [ ] from the run, never from this example):

[Topic], last [n] days · [n] posts read, [n] on topic

Themes
1. [Theme in a few words] ([n] posts)
   "[exact quote from a post]" · [where]
   [link]
2. ...

Questions people ask
- [Question] ([n] posts)

Top posts
- [Network] · @[poster]: [n] likes · [link]

Searched: [each network and term]. [n] credits used, [n] left.

Repeat

Offer a weekly version with an OpenClaw Automation (scheduled job). In repeat runs, show what changed: new themes, themes growing or fading.

Honest limits

• Data lookups return recent posts, best effort per network, not a full archive. Say "in the posts we found", never "on all of TikTok".
• No keyword search on Instagram, Facebook or LinkedIn.
• If a network returns nothing or times out, say so in the searched line.