bilibili-creator-finder
Skill4 starsChanged 3 months agoMITView on GitHub
Use this skill when the user wants to collect Bilibili/B站 candidate UP creators from keyword search results or an already-filtered Bilibili search page, deduplicate by UP主页 mid, write/update a Lark Base candidate table, and optionally enrich each UP profile with fans, likes, plays, representative videos, and public contact info. This skill does not send DMs or manage outreach status.
In this repository22 other files it gives its agents.
- aihotSkill
- douyin-scraperSkill
- feishu-cli-managerSkill
- funcinating-guidesSkill
- funcinating-newsSkill
- hot-topicsSkill
- humanizerSkill
- humanizer-zhSkill
- imap-smtp-emailSkill
- lark-base-replicaSkill
- oss-uploadSkill
- seedance-promptSkill
- shopifySkill
- skill-publisherSkill
- skill-vetterSkill
- slider-captchaSkill
- tikhub-querySkill
- tiktok-scraperSkill
- video-downloadSkill
- wechat-scraperSkill
- weibo-scraperSkill
- xiaohongshu-scraperSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
19 rules. About 0 are common, 13 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Use this skill only for candidate UP collection:
- Distinctive: line 16
Write new rows to a Lark Base candidate table.
- Distinctive: line 17
If an existing UP is found, append the keyword to the keyword field instead of creating a duplicate row.
- Distinctive: line 20
Do not use this skill for Bilibili private messages, outreach status machines, reply judging, or conversation extraction.
- Distinctive: line 24
Always use the dedicated Bilibili browser profile:
- Very specific: line 28
Prefer social-browser-profiles if available.
- Very specific: line 30
If the user has already filtered the Bilibili search page manually, use --mode current.
- Very specific: line 34
Never hardcode the user's table.
- Distinctive: line 55
Do not commit user-specific config files into a skill repo.
- Distinctive: line 69
Write to table and enrich profile data:
- Distinctive: line 122
Default to dry-run unless the user clearly asks to write.
- Distinctive: line 123
Before a large run, confirm the target table shown by the config is the intended one.
- Distinctive: line 124
Never overwrite enriched profile data by default.
- Distinctive: line 125
Write profile enrichment one record at a time so the table updates visibly.
- Distinctive: line 126
If Bilibili shows an ordered-click captcha popup, stop immediately, keep the tab open, and tell the user which UP caused the stop.
- Very specific: line 131
Before a new keyword run, confirm filter conditions with the user unless they explicitly say to use defaults.
- Very specific: line 132
Ensure the configured Bilibili browser instance is logged in and on the intended search/filter page, or choose --mode navigate.
- Distinctive: line 133
Run one dry-run for the first page and show the sample candidates.
- Very specific: line 134
Run with --execute when the target table, filter conditions, and sample look right.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Bilibili UP Candidate Scraper
Scope
Use this skill only for candidate UP collection:
- Collect UP names, UP homepage links, and source video links from Bilibili video search result pages.
- Deduplicate by normalized
https://space.bilibili.com/<mid>. - Write new rows to a Lark Base candidate table.
- If an existing UP is found, append the keyword to the keyword field instead of creating a duplicate row.
- Optionally visit UP homepages to enrich fans, likes, plays, representative videos, and public contact info.
Do not use this skill for Bilibili private messages, outreach status machines, reply judging, or conversation extraction. Those should be a separate outreach skill.
Browser Rule
Always use the dedicated Bilibili browser profile:
- Platform:
bilibili - CDP URL: configurable (recommended via
config.browser.cdpUrlor envBILIBILI_CDP_URL) - Prefer
social-browser-profilesif available; scripts fall back to the configured CDP URL.
If the user has already filtered the Bilibili search page manually, use --mode current. If the user gives keyword/sort and wants the script to open the page, use --mode navigate.
Config First
Never hardcode the user's table. Require a local config file based on templates/config.example.json.
Minimum required Lark config:
{
"lark": {
"baseToken": "...",
"tableId": "..."
}
}The config may override field names. See references/lark-schema.md for the default schema.
Recommended local config path in a project workspace:
./.bilibili-up-candidates.config.jsonDo not commit user-specific config files into a skill repo.
Main Commands
Dry-run from the currently open/filtered Bilibili search page:
node ~/.agents/skills/bilibili-creator-finder/scripts/run_candidate_pipeline.mjs \
--config ./.bilibili-up-candidates.config.json \
--keyword "示例关键词" \
--mode current \
--pages 1Write to table and enrich profile data:
node ~/.agents/skills/bilibili-creator-finder/scripts/run_candidate_pipeline.mjs \
--config ./.bilibili-up-candidates.config.json \
--keyword "示例关键词" \
--mode current \
--pages all \
--limit 50 \
--intervalMs 3000 \
--executeNavigate by keyword and sort, then write:
node ~/.agents/skills/bilibili-creator-finder/scripts/run_candidate_pipeline.mjs \
--config ./.bilibili-up-candidates.config.json \
--keyword "示例关键词" \
--sort pubdate \
--mode navigate \
--pages all \
--executeCollect search results only, without visiting UP homepages:
node ~/.agents/skills/bilibili-creator-finder/scripts/run_candidate_pipeline.mjs \
--config ./.bilibili-up-candidates.config.json \
--keyword "示例关键词" \
--mode current \
--pages all \
--enrich false \
--executeParameters
--config: Required. Local JSON config for target Lark Base and field mapping.--keyword: Required. Keyword to write into the keyword field.--mode:currentornavigate. Default iscurrent.--sort: Used only with--mode navigate. Common values:pubdate,click,totalrank.--pubtimeRange: Optional time filter. Supported values:recent6m,recent3m,recent1m.--pubtimeBegin/--pubtimeEnd: Optional Unix second timestamps for an exact Bilibili publish-time window.--pages: Page count orall. Default is1.--limit: Max UP profiles to enrich in this run. Default comes from config, usually50.--intervalMs: Delay between profile visits. Default comes from config, usually3000.--enrich: Whether to visit UP profiles after collection. Default istrue.--execute: Required for Lark writes. Without this, scripts dry-run and print samples.
Safety Rules
- Default to dry-run unless the user clearly asks to write.
- Before a large run, confirm the target table shown by the config is the intended one.
- Never overwrite enriched profile data by default. Existing UPs only get keyword union updates during search collection.
- Write profile enrichment one record at a time so the table updates visibly.
- If Bilibili shows an ordered-click captcha popup, stop immediately, keep the tab open, and tell the user which UP caused the stop.
Workflow
- Confirm or create a local config from
templates/config.example.json. - Before a new keyword run, confirm filter conditions with the user unless they explicitly say to use defaults. At minimum confirm sort (
综合排序/最新发布/最多播放), page scope (当前页/前 N 页/全部页), and whether to reuse the current manually filtered page. - Ensure the configured Bilibili browser instance is logged in and on the intended search/filter page, or choose
--mode navigate. - Run one dry-run for the first page and show the sample candidates.
- Run with
--executewhen the target table, filter conditions, and sample look right. - For long runs, use batches such as
--limit 50 --intervalMs 3000; stop only on captcha or hard error.
Files
scripts/collect_from_search.mjs: Collects UP candidates from search result pages and upserts rows/keywords.scripts/enrich_profiles.mjs: Visits UP homepages and fills profile metrics/contact fields.scripts/run_candidate_pipeline.mjs: Runs collection then optional enrichment.references/lark-schema.md: Default field mapping and required/optional fields.references/bilibili-selectors.md: Selector and captcha notes.templates/config.example.json: Copy this to a local config and fill in table info.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

