feat: add bilibili/comments, xiaohongshu/comments commands + rate-limiter plugin docs

[mikkley]

Summary

• bilibili/comments <bvid> — fetches top-level replies via /x/v2/reply/main with WBI signing. Resolves bvid → aid via /x/web-interface/view, then calls the reply API with signed: true. No DOM dependency, same pattern as bilibili/search.

• xiaohongshu/comments <note-id> — DOM extraction from the note detail page (/explore/<id>). XHS doesn't expose a public comments API, so we scroll the note container and scrape .parent-comment elements. Handles two edge cases:

  • Login-wall detection (throws AuthRequiredError)
  • Zero-like counts: XHS shows the text "赞" in the .count element when likes = 0; the extractor converts that to 0 instead of a string.

• docs/advanced/rate-limiter-plugin.md — documents the onAfterExecute hook and shows a copy-paste rate-limiter plugin that adds a random 5–30 s sleep after every browser-platform command. Useful for long-running scraping jobs. Controlled via OPENCLI_RATE_MIN, OPENCLI_RATE_MAX, and OPENCLI_NO_RATE=1 (disable for local dev).

Test plan

• npm test — all 294 existing tests pass; 10 new tests added (5 per command)
• npx tsc --noEmit — no type errors
• Manual: opencli bilibili comments BV1WtAGzYEBm --limit 5 returns ranked comment rows
• Manual: opencli xiaohongshu comments <note-id> --limit 5 returns comment rows when logged in
• Manual: rate-limiter local install prints [rate-limiter] xiaohongshu: sleeping Xs to stderr

🤖 Generated with Claude Code

[Astro-Han]

Both comments commands follow existing patterns nicely: bilibili reuses WBI signing, xiaohongshu goes with DOM extraction. A few things I noticed:

bilibili/comments error handling

The bvid resolution failure throws a bare Error instead of a CliError subclass (existing subtitle.ts uses CliError). Also, responses from /x/web-interface/view and /x/v2/reply/main don't check payload.code, so rate-limiting or auth failures would silently return an empty array instead of a clear error.

xiaohongshu/comments can't consume search output

xiaohongshu/search outputs URLs like /search_result/...?xsec_token=..., but comments only recognizes /explore/<id> and /note/<id>. Passing a search result URL to comments would fail to parse.

Output field leak

Each comment object includes loginWall: false, which leaks into --format json output and doesn't match the declared columns.

Empty result semantics

xiaohongshu throws EmptyResultError on zero comments, while bilibili returns an empty array. Might be worth aligning.