XHS Account Research

Follow shared release-shell rules in:

• postplus-shared release-shell rules

Legacy alias: xhs-account-research.

Use this skill when the user wants to:

• compare one or more Xiaohongshu accounts
• build a shortlist of benchmarkable accounts from known profile URLs
• understand what an account usually posts, what title patterns repeat, and which notes perform best
• produce account-level snapshots for partner or competitor review

Read these references before implementation:

• skills/20-research/xhs-account-research/references/actor-selection.md
• skills/20-research/xhs-account-research/references/normalized-schema.md

Default posture

Start from validated recent-post evidence.

Default actor:

• rednote-xiaohongshu-user-posts-scraper

Do not default to rednote-xiaohongshu-profile-scraper. Treat it as non-default until it returns usable profileData on the released shell.

What this skill is for

• recent-post sampling from known profile URLs
• account-level aggregation from post evidence
• ranking accounts by observed post performance and thematic fit
• producing account snapshots with top notes and repeated title patterns

What this skill is not for

• keyword discovery
• comment mining
• follower counts
• bio extraction
• posting cadence claims when no publish timestamps are available

Failure posture

• fail if no profileUrls or profileIds are provided
• fail if the validated actor returns zero posts
• fail if account aggregation cannot produce any top note URL
• report missing profile-level fields explicitly instead of inventing them

Release-Shell Execution Contract

• keep account briefs, actor inputs, raw datasets, normalized datasets, rankings, and analysis outputs under <work-folder>/.postplus/xiaohongshu-account-research/
• keep only final user-facing account shortlists or reports outside .postplus/
• start with a bounded first pass, usually 1-3 accounts and 8-15 recent posts per account
• if hosted capability is unavailable, unauthorized, or returns a stable network error, stop immediately instead of switching to ad hoc shell glue

Main scripts

• scripts/build_xhs_account_actor_input.mjs
• scripts/normalize_xhs_account_dataset.mjs
• scripts/rank_xhs_accounts.mjs
• scripts/analyze_xhs_accounts.mjs

Use the shared collection runner for actor calls:

• ${CLAUDE_SKILL_DIR}/_postplus_shared/00-core/shared-collection/scripts/collection_actor_run.mjs

Minimal workflow

node ${CLAUDE_SKILL_DIR}/scripts/build_xhs_account_actor_input.mjs \
  --brief <work-folder>/.postplus/xhs-account-brief.json \
  --output <work-folder>/.postplus/xhs-account-actor-input.json

node ${CLAUDE_SKILL_DIR}/_postplus_shared/00-core/shared-collection/scripts/collection_actor_run.mjs \
  --collection-key xhs-account-posts \
  --input <work-folder>/.postplus/xhs-account-actor-input.json \
  --output <work-folder>/.postplus/xhs-account-raw.json

node ${CLAUDE_SKILL_DIR}/scripts/normalize_xhs_account_dataset.mjs \
  --input <work-folder>/.postplus/xhs-account-raw.json \
  --output <work-folder>/.postplus/xhs-account-normalized.json

node ${CLAUDE_SKILL_DIR}/scripts/rank_xhs_accounts.mjs \
  --input <work-folder>/.postplus/xhs-account-normalized.json \
  --theme "职场,打工人" \
  --output <work-folder>/.postplus/xhs-account-ranking.json

node ${CLAUDE_SKILL_DIR}/scripts/analyze_xhs_accounts.mjs \
  --input <work-folder>/.postplus/xhs-account-normalized.json \
  --output <work-folder>/.postplus/xhs-account-analysis.json

Good output

Return:

• account shortlist
• top note URLs per account
• median and max like signal
• content-type mix
• repeated title-pattern families
• cover-shape mix
• explicit data-quality warnings for unavailable profile-level fields