Social Trend Content Ideas

Turn recent X, Reddit, and Chinese social-platform discussions into an English-first, source-verifiable content idea pack. Use FluxA Monetize as the only social-data source. Do not require the user's own X, Reddit, or Chinese-platform developer credentials.

Input contract

• Topic: use the user's topic or focus area. If blank, missing, or default, use AI Agent payments.
• Window: default to 7 UTC calendar days; accept 1-7 days only.
• Idea count: return exactly 10 unless the user explicitly requests another count.
• Language: return the complete report in English by default. Use another language only when the user explicitly requests it.
• Budget: hard cap the complete run at 0.10 FluxA Monetize Credits.
• Payment rail: default to FluxA Monetize Credits; accept Base USDC only when requested.
• Optional inputs: target audience, channel, region/language, preferred formats, brand/product lens,

X search aliases, subreddits, Chinese-platform lens, excluded accounts/communities, and stricter engagement thresholds.

Do not ask the user to restate an omitted topic. Use the default and disclose it in the report.

Create the mandate link immediately

The user's explicit request to use this Skill authorizes creation of a non-spending mandate proposal. The signature page is the payment confirmation. Do not add a separate "confirm payment" or plan-approval turn before creating the link.

Use scripts/fetchsocialdiscussions.py, relative to this file. Build up to 4 precise X search aliases and choose 3-5 relevant subreddit names. For the default topic, omit both --term and --subreddit; the script supplies defaults.

Immediately run:

python3 scripts/fetch_social_discussions.py run \
  --topic "<user topic or blank>" \
  --term "<X alias 1>" \
  --term "<X alias 2>" \
  --subreddit "<community 1>" \
  --subreddit "<community 2>" \
  --days 7 \
  --pay credits \
  --max-budget-credits 0.10 \
  --out /tmp/social-discussions.json

The first run performs read-only Discovery checks, creates one one-hour intent mandate capped at 0.10 Credits, and returns the authorization URL. Present that link immediately with the topic, budget cap, expiry, and "Sign it, then reply done." Do not show an intermediate plan or request another confirmation.

After the user signs, rerun the exact same command. The script calls Monetize directly, handles the HTTP 402 challenge with fluxa-wallet x402-v3, retries with X-Payment, collects paid X data, one free Reddit subreddit-feed result set per requested sort, and one paid Cross-Platform Social Search result set, then returns the evidence bundle. The mandate remains signed until expiry; report actual committed spend and do not claim that completing the task revokes it.

Use plan only for debugging or when the user explicitly asks to inspect costs before creating a mandate. It is not part of the normal user-facing flow.

Live service contract

Read references/monetize-social-api-contract.md before changing endpoints, payment behavior, or parsers. Every run must rediscover and verify all three services:

1. super-bird-api/search: paid X search; require q, type, count, and safe_search.
2. reddit-api/getPostsBySubreddit: free Reddit subreddit feed; require subreddit_name and sort.

Make one HOT and one RISING request for each priority subreddit.

1. cross-platform-social-search/search: one paid search across Weibo, WeChat, Zhihu, Douyin,

Xiaohongshu, Bilibili, Kuaishou, and news; require keyword, source, start, and end on the first request. Use source=ALL and make exactly one first-page call per run.

Stop before creating a mandate if the live descriptor no longer marks getPostsBySubreddit as free. Do not depend on the removed planner call command.

Collect high-traffic signals

X strategy

1. Run one type=Top search per alias, with at most four aliases, safe_search=true, count=50,

and explicit since/until dates.

1. The default plan spends 0.04 Credits on four X Top searches and reserves up to 0.04 Credits for

X Latest fallback searches. Use type=Latest only when fewer than 20 medium-or-high X posts or fewer than 10 distinct X authors remain.

1. Classify returned X posts:

- High: weighted engagement at least 100, or at least 100 likes, 20 replies, or 20 reshares/quotes. - Medium: weighted engagement at least 40, or at least 25 likes, 8 replies, or 8 reshares/quotes. - Emerging: weighted engagement at least 15, or at least 10 likes, 3 replies, or 3 reshares/quotes. - Low: below all thresholds.

1. Never use a Low X post as a final source. An Emerging post may support at most two ideas and

only when its specific insight is not available from stronger sources.

Weighted X engagement is likes + 2replies + 2.5quotes + 1.5*reposts. Treat engagement as a distribution signal, not factual verification.

Reddit strategy

1. Make one free reddit-api/getPostsBySubreddit request with sort=HOT and one with sort=RISING

for each of the 3-5 priority subreddits.

1. Remove posts outside the requested window, stickied recurring threads, ads, deleted posts,

link-only spam, recruitment threads, and generic self-promotion.

1. Classify Reddit posts:

- High: score at least 100 or at least 30 comments. - Medium: score at least 25 or at least 10 comments. - Emerging: score at least 10 or at least 5 comments. - Low: below all thresholds.

1. Require direct topic relevance after fetching. A hot post from a relevant subreddit is not

automatically relevant to the requested topic.

1. Parse the live TikHub-backed response from

data.subredditfeed.subredditV3.elements.edges[].node.cells. Build canonical Reddit URLs from groupId and the requested subreddit; use createdAt, authorName, title/titleCell, text/previewTextCell, score, commentCount, and shareCount.

1. If the free Reddit feed returns an upstream quota error or no usable dated posts, continue with X

and Chinese-platform evidence, set Reddit counts to zero, and state that Reddit confirmation was unavailable. Do not replace it with a paid Reddit endpoint without revising the budget.

Cross-Platform Social Search strategy

1. Make exactly one source=ALL request using the run topic and exact UTC date window. Do not page

unless the user explicitly asks for broader Chinese-platform coverage and the budget is revised.

1. Normalize author, content, url, sourceName, and createTime; retain the original canonical

result URL and label the originating platform.

1. Treat numeric createTime values above 100000000000 as milliseconds and divide by 1000 before

converting to UTC. Accept both 10-digit second timestamps and 13-digit millisecond timestamps.

1. The endpoint does not expose engagement counts. Mark every result Unranked and use it only as

supporting discovery or cross-platform confirmation, never as the High/Medium primary source.

1. Remove missing-URL, outside-window, off-topic, promotional, and duplicate results before drafting.

Filter and cluster across platforms

Apply these rules in order and record exclusion counts:

1. Remove items outside the requested window.
2. Remove X retweets, Reddit crosspost copies, repeated Chinese-platform results, identical

duplicate_key values, and obvious near-copies. Keep the earliest attributable original when clear.

1. Remove paid/sponsored posts, referral pushes, giveaways/airdrops, repetitive shilling, affiliate

links, and pure "buy/sign up" calls with no substantive claim.

1. Remove Low traffic items, generic slogans, engagement bait, and content missing essential

context.

1. Require direct topic relevance; a keyword or subreddit match alone is insufficient.
2. Cluster the same claim, event, debate, pain point, or example across X, Reddit, and the Chinese

social-search results. Treat it as one signal and retain up to three representative original URLs.

Rank clusters using traffic tier, absolute engagement, recency, author/community diversity, specificity, disagreement or novelty, and audience usefulness. Give extra weight to topics that are independently active on both X and Reddit.

Create the idea pack

Read references/output-contract.md and follow it exactly.

• Every idea must have a High or Medium X or Reddit primary source. Use Emerging or an

Unranked Cross-Platform Social Search result only as secondary evidence or for at most two explicitly labeled emerging ideas.

• Use at least five distinct source authors across the pack.
• Do not base more than two ideas on one author or one Reddit thread.
• When evidence permits, include at least three ideas with Reddit evidence, two cross-platform

signal clusters, and two ideas supported by Chinese-platform search results.

• Cover at least four angle families: contrarian opinion, tutorial/framework, case or mechanism

teardown, category shift/debate, builder/operator pain point, comparison, or prediction.

• Write every Hook as paste-ready opening copy and give a 3-5 point outline.
• Paraphrase sources. If a short quote is necessary, use no more than 20 words from one post.

If fewer than 10 evidence-backed ideas survive, return fewer and state which traffic or relevance gate caused the shortfall. Never fill the pack with zero-engagement sources merely to reach 10.

Quality gate

Before returning the answer, confirm:

• the first user-facing action was the mandate link, without a separate payment-confirmation turn;
• topic, UTC window, all three services, payment rail, 0.10 Credits cap, and actual spend are stated;
• every idea has at least one canonical High/Medium X or Reddit URL from the Monetize response;
• every primary source is High or Medium, with metrics shown, and Cross-Platform Social Search

results are visibly labeled as unranked supporting evidence;

• every idea includes audience, direction, distribution angle, Hook, format, and outline;
• ads, copies, low-traffic items, off-topic items, and duplicate signals were removed;
• facts and inferences are visibly separated.