Rednote research workflow

RedNote Research Workflow

Use this skill when RedNote/Xiaohongshu can help you understand examples, user language, informal practices, creator communities, or trend signals.

Treat the platform as social evidence. It can suggest what to look at next. It should not be the final authority for rules, dates, eligibility, official templates, prices, medical facts, legal claims, or academic claims.

Good use cases

• finding how students, creators, or practitioners describe a topic
• collecting examples of portfolios, reports, resumes, competition materials, or project showcases
• discovering creator accounts, keywords, hashtags, and adjacent topics
• comparing wording, formats, and practical expectations around a domain
• finding leads that should later be verified against official or primary sources

Do not use this skill for account growth, engagement automation, scraping at scale, or private account activity.

Read-only rule

Default to read-only research.

Do not publish, like, collect, comment, follow, message, delete, or batch-operate unless the user explicitly asks and the local environment supports it safely. Even then, confirm the exact action before doing it.

Workflow

1. Define the research question

Write one sentence:

• what you are trying to learn
• who the observed community is
• what decision the findings should support

Examples:

• "Find how applicants present undergraduate research experience in portfolio posts."
• "Collect visual and wording references for AI competition project showcases."
• "Identify common mistakes people mention when preparing a design competition board."

2. Build keyword groups

Use 3 to 10 queries. Mix:

• direct topic words
• audience words
• output format words
• pain-point words
• adjacent community words

Example groups:

• research: paper reproduction, literature reading, research tools, experiment design
• coursework: course report, presentation deck, project showcase, defense slides
• competitions: innovation competition, design competition, data contest, project proposal
• career or school materials: resume, portfolio, advisor email, personal statement

Localize keywords to the actual language used by the platform community.

3. Screen results

Prefer notes with:

• concrete title
• useful images, screenshots, templates, code, tables, or before/after examples
• comments that reveal real questions or objections
• author profile relevant to the task
• enough engagement to suggest visibility, without treating engagement as truth
• links or clues pointing to official sources

Downrank notes that are pure inspiration, generic advice, copied lists, unsupported claims, or obvious ads.

4. Read shortlisted notes

For each useful note, keep:

• URL
• title
• author or account label
• retrieval date
• keyword used
• visible metrics if available
• what the note is useful for
• what needs verification

Do not copy large chunks of platform content into a final report. Summarize the relevant signal and keep links for traceability.

5. Separate signal from fact

Use RedNote/Xiaohongshu for:

• leads
• examples
• phrasing
• community pain points
• creator discovery
• visual conventions
• trend hints

Verify elsewhere:

• competition rules
• deadlines
• submission formats
• eligibility
• school or company policies
• medical, legal, or financial claims
• academic facts and citations

Read references/evidence-quality.md when a finding may affect a formal decision.

Output

Default to a concise source-oriented table:

Use	Keyword or note	Signal	Evidence	Needs verification

Then add:

• top 3 useful patterns
• top 3 weak or risky signals
• one recommended next verification step

Boundaries

• Do not expose session data or account state.
• Do not save private platform data into a public repository.
• Do not present social metrics as stable.
• Do not imply official endorsement from social posts.
• Do not use platform content as the sole source for formal claims.