SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsfastmoss-creatives

fastmoss-creatives

Data & analytics

View repository

Updated 10-16 08:57

About this skill

fastmoss-creatives

End-to-end FastMoss creative material ranking scraper + analyzer. Uses kimi-webbridge to drive the user's logged-in browser, extracts rows into CSVs, and ships an aggregation script + Markdown report template that produces an analysis report. Scope: the Videos & Materials section of fastmoss.com (3 table-based rankings; the AI Sales Video Ranking is card-based and noted below). For products, shops, creators, ads, or live streams, use the corresponding sister skill.

Prerequisites

  1. Kimi WebBridge daemon healthy:

    ~/.kimi-webbridge/bin/kimi-webbridge status
# expect: {"running": true, "extension_connected": true}

    If not healthy: invoke Skill(kimi-webbridge) and follow references/operations.md.

  2. User is logged in to fastmoss.com in their browser (this skill reuses their session — no auth).

  3. Read references/environment.md before writing new shell or evaluate JS.

Creative ranking URLs

Ranking key Chinese URL Schema
video Popular Videos https://www.fastmoss.com/zh/media-source/video table
song Popular Music https://www.fastmoss.com/zh/media-source/song table
hashtag Popular Hashtags https://www.fastmoss.com/zh/media-source/hashtag table
top-ai-videos AI Sales Video Ranking https://www.fastmoss.com/zh/media-source/top-ai-videos cards (not scripted)

Use /zh/ (Chinese) locale. Site is a NUXT/Vue SPA — always wait for hydration (--nav-sleep 6+).

Per-ranking column schemas

Ranking Header columns
video Video Content, Creator Info, Creator Followers, Publish Time, Views, Likes, Engagement Rate
song Music, Creator, Views
hashtag Hashtag, Preview, Impressions, Related Video Count

The Video Content/Music/Hashtag columns are multi-line text; the scraper pre-parses the primary entity name into entity_name and keeps the raw cell under its original header.

Quick start

# Step 1: Scrape one ranking (Top N pages)
python <path-to-skill>/scripts/creative_scraper.py \
    --ranking video --pages 5 --out <out-dir>/videos.csv
python <path-to-skill>/scripts/creative_scraper.py \
    --ranking song --pages 3 --out <out-dir>/songs.csv
python <path-to-skill>/scripts/creative_scraper.py \
    --ranking hashtag --pages 3 --out <out-dir>/hashtags.csv

# Step 2: Multi-country comparison on one ranking
python <path-to-skill>/scripts/creative_filtered.py \
    --ranking video --country 美国,印度尼西亚,泰国,马来西亚 \
    --pages 3 \
    --out <out-dir>/videos_by_country.csv

# Step 3: Aggregate the per-ranking CSVs into a Markdown report
python <path-to-skill>/scripts/analyze.py \
    --video <out-dir>/videos.csv \
    --song <out-dir>/songs.csv \
    --hashtag <out-dir>/hashtags.csv \
    --out-md <report>.md

<path-to-skill> is wherever this skill is installed (typically .claude/skills/fastmoss-creatives inside the project, or ~/.claude/skills/fastmoss-creatives for a user-level install). Run any script with --help for the full arg list.

Scripts

Script Purpose
scripts/creative_scraper.py Top-N paginating scraper, any of the 3 table rankings (schema auto-detected from <thead>)
scripts/creative_filtered.py Multi-country filter + paginate
scripts/analyze.py Aggregate CSVs → crosstabs (stdout) or full Markdown report (template-driven)

References (read on demand)

File When to read
references/environment.md Before writing any new shell or evaluate JS — Windows/bash/jq/heredoc rules (mirrors fastmoss-products' version)
references/analysis_recipe.md The Markdown template analyze.py fills. Read to understand the report structure or to extend it

Filter label vocabularies

Countries (16 markets)

--country <中文,comma-separated>: 全部, 美国, 印度尼西亚, 英国, 越南, 泰国, 马来西亚, 菲律宾, 西班牙, 墨西哥, 德国, 法国, 意大利, 巴西, 日本, 新加坡.

Same vocabulary as fastmoss-products / fastmoss-creators.

Critical gotchas (each cost real time on first run)

Full details in references/environment.md. Summary:

  1. @e refs from snapshot go stale after every navigation → "No node with given id found". Bundled scripts click by text via evaluate, never @e refs.
  2. Bash heredoc + JS regex = "Invalid regular expression: missing /" — heredoc eats a backslash layer. Use .split('\n') over .replace(/\n+/g, ...).
  3. <thead> headers may report empty via innerText on first render. The bundled EXTRACT_JS reads innerText || textContent and falls back to the title attribute.
  4. SPA hydration lag — pages may need slightly longer waits than product pages (6s default). Too fast → 0 rows.
  5. Schema differs between rankings — creative_scraper.py reads <thead> dynamically. Don't assume a fixed schema in downstream code; iterate over dict.keys().
  6. jq may not be installed — decode screenshots manually with python -c + base64.
  7. Python defaults to a locale-specific encoding on Windows — always open(..., encoding='utf-8-sig') when reading the produced CSVs.
  8. The AI Sales Video Ranking has no <table> — it's a card grid. Not covered by the bundled scraper. Write a custom evaluate per references/environment.md if needed.

CSV output schema

Base columns (always present):

page, ranking, rank, entity_name, country

Plus one column per <thead> header (Chinese, kept as-is). All CSVs are utf-8-sig (BOM) for Excel.

When using creative_filtered.py, one extra column is prepended:

filter_country

Session hygiene

  • Default session "fastmoss-creatives" for all calls (isolated from sister skills' sessions). Override via --session if needed.
  • At task end: curl -d '{"action":"close_session","args":{},"session":"fastmoss-creatives"}' http://127.0.0.1:10086/command.
  • All bundled scripts use newTab:True on first navigate so they work whether or not the session pre-exists.