hc-tec/sourcerelay-agentkit/collect-bilibili
Claude Code and agent skill by hc-tec/sourcerelay-agentkit
- community
Checking that the SKILL.md file is still in the repository.
Collect public Bilibili information through the registered typed Collector MCP Tools, including bounded interpretation of subtitle fields when the live Core Artifact contract exposes them. Use for Bilibili keyword search, video detail, UP account profile,投稿 inventory, dynamics, collection or series overview/detail, danmaku, public discussion, or a request that needs the current subtitle capability boundary. Use together with use-collector-mcp whenever an Agent must choose a Bilibili capability, construct its bounded input, sequence related reads, or interpret platform-specific prerequisites without arbitrary browser control.
Install
npx vskill@latest install hc-tec/sourcerelay-agentkit/collect-bilibiliSKILL.md
Collect Bilibili
Use only the Bilibili Tools currently present in MCP. Apply $use-collector-mcp for compatibility, binding selection, idempotency, Operation monitoring, and bounded Artifact reads.
Choose the smallest capability
- Use
collector_bilibili_native_searchfor one bounded relevance search. - Use
collector_bilibili_native_search_batchonly when the request genuinely needs the registered
broader multi-page search budget.
- Use
collector_bilibili_video_detailfor one known canonical BV video. - Use
collector_bilibili_account_profilefor public UP identity and profile facts. - Use
collector_bilibili_account_inventoryfor public投稿 inventory. - Use
collector_bilibili_dynamicfor the registered bounded public dynamic feed. - Use
collector_bilibili_collection_series_overviewbefore requesting an unknown series or season. - Use
collector_bilibili_collection_series_detailonly with a stable ID and type obtained from
overview evidence.
- Use
collector_bilibili_discussionfor public video discussion. - Use
collector_bilibili_danmakuonly when danmaku is part of the user's requested evidence. - For subtitle requests, first read
collector://capabilities. Use a standalone transcript Tool only
if the live catalog reports bilibili.transcript as direct_ready; do not invent a Tool from the existence of Core research code or a historical reconnaissance report.
Read references/capabilities.md before constructing Tool arguments or chaining more than one capability.
Keep inputs canonical
Use canonical public URLs without query or fragment:
Prefer URLs and stable IDs returned by a prior Artifact. Do not guess a BV ID, UID, series ID, or season ID from display text. Let the Tool/Core schema reject invalid inputs; never work around it with navigation, selectors, scripts, tab IDs, or API simulation.
For collector_bilibili_account_inventory, choose collector_work_tab normally. Choose user_selected_tab only when the current MCP binding has a valid user-created short lease; never provide or infer a tab identity.
Sequence explicit operations
For search-to-video research:
- Run one search Tool with a new request ID.
- Read its terminal Artifact and rank results in caller-owned reasoning.
- Select only relevant canonical BV URLs.
- Run video detail for selected results with one new request ID per video.
- Add discussion or danmaku only when requested or when a stated evidence gap requires it.
For account research, run profile, inventory, dynamics, and series overview as separate visible operations according to the user's requested breadth. Run series detail only for selected overview items. Do not hide a loop, pagination policy, or account-wide archive inside the Skill.
Preserve platform facts
- Treat charge-only, unavailable, login, risk, partial, empty, and failed outcomes as distinct facts.
- Do not infer standalone subtitle capability from a video-detail result. The live capability catalog
remains the source of truth, and trusted_interaction_migration_required means the MCP surface must stop without submitting a platform Operation.
- If a direct-ready video-detail Artifact includes a bounded
subtitleprojection, preserve its
available, language, panelVisible, segmentCount, partial, segment timing, and content fields exactly; do not claim that an empty projection proves the user is logged out.
- The Bilibili subtitle hover implementation is owned by Core's paired extension. This Skill never
sends hover/click/CDP commands and never falls back to browser-control primitives.
- Do not call every Tool merely because it exists.
- Keep result provenance through Operation and Artifact identities and hashes.
- Stop when the requested coverage is met, the explicit budget is exhausted, or Core reports a safety
terminal.
All Tools perform public reads through the user's paired browser. Do not like, follow, favorite, comment, message, publish, delete, bypass paid content, or evade platform controls.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/collect-bilibili/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/hc-tec/sourcerelay-agentkit/collect-bilibili)