Voice of Customer Content Engine

You have access to the SnitchFeed MCP.

Build me a content plan from what people are actually saying, not from what a keyword tool guesses they search.

Ground rule: every quote and every example must trace back to a real mention with a link. Do not invent customer stories, statistics, or examples.

1. Call snitchfeed_get_org_context so you know what we sell and who we sell to.

2. Call snitchfeed_list_mentions with:
   - include_intents: ["pain_point", "content_opportunity", "recommendation_request", "comparison", "feature_request"]
   - date_from: 90 days ago
   Paginate until you have them all.

3. Call snitchfeed_query_analytics twice for context:
   - metric "mentions", group_by "keyword", same date range, to see which terms carry the volume
   - metric "mentions", group_by "subreddit", to see where the conversation actually lives

4. Cluster the mentions into recurring problems. For each cluster capture the exact phrasing people use, including the blunt, unpolished, non-marketing wording. That phrasing is the asset, so quote it verbatim rather than tidying it up.

5. Rank the clusters on three things: how often they appear, how much buying intent sits next to them (look at whether buying_intent or recommendation_request tags show up in the same cluster), and whether we can credibly answer given our positioning.

6. For the top 8 clusters give me:
   - the problem in the customer's own words
   - a working title that uses their phrasing, not ours
   - the angle, in one sentence
   - the three questions the piece has to answer to be worth reading
   - 2 to 3 real mention permalinks as source material
   - the format that fits best: blog post, landing page, or social post

7. Finish with a list of phrases we should stop using because nobody outside the company talks that way, based on the gap between our positioning language and theirs.

If I have an SEO data tool connected, cross-check the top phrasings for search volume and tell me which are worth a page versus a post.