Social Media Research Ops

Social Media Research Ops

Overview

Use this skill to run a repeatable social media research process: define the research target, browse real platform samples, record posts and comments, infer platform-specific methods, and write structured outputs into Obsidian.

Use the Obsidian vault path provided by the user. If no path is provided, inspect the available workspace or ask only when writing files is required and no reasonable location can be inferred.

Operating Rules

• Browse slowly and behave like a reader when using logged-in social platforms.
• Do not like, follow, comment, post, DM, bookmark, collect, change settings, or otherwise mutate social accounts.
• If a platform shows verification, captcha, login, rate limit, or unusual security prompt, stop that platform and tell the user.
• Separate verified observations from inferred conclusions.
• Prioritize real posts, comment sections, search suggestions, profile positioning, and repeated patterns over generic marketing theory.
• Write durable findings to Obsidian whenever the task is research-heavy or the user asks for reusable output.

Workflow

1. Set Research Scope

Clarify or infer:

• Product or project being promoted.
• Target users and regions.
• Platforms to study.
• Competitor categories.
• Desired output: strategy, platform playbook, post ideas, account matrix, daily report, or all of these.

If the user gives no platform priority, start with:

1. X/Twitter for global AI/video/creator ecosystem and model discourse.
2. Xiaohongshu for Chinese search, tutorials, save-worthy content, and comment demand.
3. TikTok for short-form visual hooks.
4. Reddit for honest demand research and anti-promo constraints.

2. Build The Sample Map

Collect samples across three layers:

• Direct competitors: same product category or same user job.
• Adjacent accounts: creator tools, AI video generators, model labs, workflow creators, tool curators.
• Demand communities: Reddit threads, search result pages, comment sections, Q&A posts.

For each account or query, record:

• Platform, account/query, URL, date accessed.
• Positioning and audience.
• Follower or visible credibility signal.
• Representative post titles/topics.
• Engagement signals.
• Repeated content formats.
• Comment-section questions and objections.

3. Read Posts And Comments

For high-value posts, capture:

• Hook/title/cover.
• Content structure.
• Offer or implicit promise.
• Tool/model/workflow mentioned.
• Engagement numbers when visible.
• Comment questions, objections, buying signals, and confusion.
• What the creator replies to and ignores.

Do not over-index on one viral post. Look for repeated needs across accounts, searches, and comments.

4. Infer Platform Logic

Analyze each platform separately before making cross-platform conclusions.

For X/Twitter:

• Look for model launches, creator workflows, benchmark demos, quote/reply dynamics, credibility-building threads, and founder/creator conversations.
• Favor tested judgment over plain links or news reposts.

For Xiaohongshu:

• Look for search terms, related searches, high-save tutorials, workflow diagrams, tool selection tables, price/cost discussions, and comments asking how, where, free, member, prompt, queue, and can it do my scene.
• Favor scenario-led, save-worthy content over abstract product claims.

For TikTok:

• Look for first-second hook, visual result, on-screen text, save/share triggers, comment questions, and pinned tool/cost details.
• Favor short demos and same-input comparisons.

For Reddit:

• Read community rules before proposing posts.
• Treat Reddit as demand research and transparent comparison, not hard promotion.

5. Convert Findings Into Strategy

Produce strategy in this order:

1. Core positioning.
2. Platform-specific role.
3. Content pillars.
4. First 10-30 post ideas.
5. Account matrix if useful.
6. Comment/reply strategy.
7. Risks and anti-patterns.
8. Next research breakpoint.

For AI video aggregation products, a useful default hypothesis is:

Do not lead with "tool directory"; lead with "AI video workflow researcher / tool selection advisor / cost and limitation guide".

Validate or revise this hypothesis against new data.

Obsidian Outputs

Use Obsidian links and short, navigable files. Prefer a main index plus platform-specific notes instead of one giant document.

For an existing vault, preserve its established grouping. A useful structure has a social media operations folder with platform folders for X, Xiaohongshu, Reddit, and TikTok, plus overview, competitor library, execution plan, and raw research or method folders.

Recommended structure:

• social-media-index.md
• social-media-x-deep-dive-stage-n.md
• social-media-xiaohongshu-deep-dive-stage-n.md
• social-media-tiktok-observations-stage-n.md
• social-media-reddit-observations-stage-n.md
• social-media-stage-conclusion-yyyy-mm-dd.md
• social-media-competitor-account-library.md
• daily-reports/yyyy-mm-dd-work-report.md
• memory/ai-collaboration-memory-project-and-working-style.md

When producing substantial outputs, read references/templates.md for reusable note templates.

Daily Memory And Reports

When the user asks for daily work logging, end-of-day reports, project memory, or continuity:

• Write a daily report under the daily reports folder.
• Update long-term memory only with durable facts: project state, preferences, decisions, constraints, repeated workflows.
• Do not store one-off browsing noise in long-term memory.

Daily reports should include:

• Work completed.
• Files created or updated.
• Key judgments.
• User preferences learned.
• Unfinished items.
• Next continuation point.
• Risks or account/platform issues.

Quality Bar

Before finishing, check:

• Each platform has its own conclusion.
• Comment-section demand is captured, not just post topics.
• Recommendations are actionable enough to publish from.
• The user can resume from the next breakpoint.
• Obsidian notes are easy to navigate from a main index.