fix(scrape): keep every page when multi-URL filenames collide

[breken-ai]

What's wrong

firecrawl scrape with several URLs saves each page to .firecrawl/<host>-<path>.md (urlToFilename). The name drops the query string and turns / into -, so different URLs can map to the same file:

$ firecrawl scrape "https://example.com/list?page=1" "https://example.com/list?page=2" \
    https://example.com/a/b https://example.com/a-b
Scraping 4 URLs...
[1/4] Saved: .firecrawl/example.com-list.md
[2/4] Saved: .firecrawl/example.com-list.md
[3/4] Saved: .firecrawl/example.com-a-b.md
[4/4] Saved: .firecrawl/example.com-a-b.md

Completed: 4/4 succeeded
$ ls .firecrawl
example.com-a-b.md  example.com-list.md

The run reports all four pages as saved, but only two files exist. Whichever request finished last overwrote the other page. Paginated URLs such as ?page=N or ?id=N are a common reason to pass several URLs to one command, and they always hit this.

Fix

Filenames are chosen before the requests start, in the order the URLs were given. If a name is already used in the batch, the later URL gets -2, -3, and so on (example.com-list-2.md). URLs that don't collide keep the same names as before, so existing scripts that read .firecrawl/<host>-<path>.md are unaffected.

Tests

New src/__tests__/commands/multi-scrape-filenames.test.ts (mocked client and fs):

• the four colliding URLs above are written to four distinct files, each with its own content. This fails on main, where only two files are written.
• URLs that don't collide keep their plain names.

pnpm run format:check, pnpm run type-check, pnpm run build and pnpm test (642 tests) pass. Under heavy local load, some subprocess tests in alexandria-beta.test.ts timed out on both main and this branch. They pass with --maxWorkers=4.

---

This PR was prepared with an AI coding agent (Claude). I reproduced the bug and checked the fix and tests before opening it.

---

Summary by cubic

Fixes firecrawl scrape silently losing pages when URLs bundled in one run collide on their output filename. URLs that differ only in their query string (?page=1 vs ?page=2) or in / versus - (/a/b vs /a-b) mapped to the same .firecrawl/<host>-<path>.md file, and the last request to finish overwrote the rest while the run still reported every URL as saved.

• Filenames are assigned up front in URL order, with a -2, -3, ... suffix for names already used in the batch. Plain names are reserved first so a suffixed name can't take another URL's plain name, and names are compared case-insensitively to match macOS and Windows.
• Non-colliding URLs keep their existing names, so scripts that read .firecrawl/<host>-<path>.md are unaffected.
• Adds tests covering colliding, non-colliding, plain-name-reserved, and case-only-colliding batches.

^{Written for commit db4636e. Summary will update on new commits.}

[cubic-dev-ai]

All reported issues were addressed across 2 files

<sub>Shadow auto-approve: would not auto-approve because issues were found.

Fix all with cubic | Re-trigger cubic</sub>

[cubic-dev-ai]

0 issues found across 2 files (changes from recent commits).

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

<sub>Shadow auto-approve: would auto-approve. Fixes multi-URL scrape losing pages when URLs collide on output filenames by assigning unique suffixed names up front, preserving plain names for non-colliding URLs, with tests covering colliding and edge cases.

Re-trigger cubic</sub>