Skip to content

docs(developer-index): document the web: id prefix - #218

Open
gaurav-init wants to merge 2 commits into
firecrawl:mainfrom
gaurav-init:fix/213-document-web-prefix-in-developer-skill
Open

docs(developer-index): document the web: id prefix#218
gaurav-init wants to merge 2 commits into
firecrawl:mainfrom
gaurav-init:fix/213-document-web-prefix-in-developer-skill

Conversation

@gaurav-init

@gaurav-init gaurav-init commented Sep 5, 2026

Copy link
Copy Markdown

Summary

/v2/search/developer returns a fifth id prefix, web:, alongside the four curated kinds documented in the skill. Unscoped queries can return a majority of web: results, which don't come from the index. Following the skill's existing "scope last, not first" guidance and satisfying its "quote the passage, cite the url" rule then treats an open web page as a curated primary source.

This updates the enumerated prefix list in skills/firecrawl-developer-index/SKILL.md to mention web:, describe it as an open web page rather than a curated artifact, and note the workaround (pass types=["doc","issue","pull_request","readme"] to keep only curated results; types=["web"] is rejected with a 400).

Minimal single-line edit — no changes to principles, filters, or examples.

Related

Fixes #213

Test plan

  • pnpm format:check — clean
  • pnpm type-check — clean
  • pnpm test — 426 passed / 7 skipped, no regressions
  • Skill description still under Claude Code's 1024-char frontmatter limit (205 chars)

Summary by cubic

Documents the web: id prefix that /v2/search/developer returns alongside the four curated index kinds. The skill previously listed only doc:, issue:, pull_request:, and readme:, so an agent following the "scope last" guidance could cite an open web page as if it were a curated primary source. The update marks web: results as open web pages and clarifies in the Filters section that unset types returns them too — passing types=["doc","issue","pull_request","readme"] is the only way to exclude them, while types=["web"] alone is rejected with a 400.

Fixes #213.

Written for commit c8e383d. Summary will update on new commits.

Review in cubic

Search API replies to `/v2/search/developer` include a fifth id
prefix — `web:` — for open web pages returned alongside the four
curated kinds (`doc:`, `issue:`, `pull_request:`, `readme:`). The
skill enumerated only the curated four, so an agent following the
"scope last" guidance and satisfying the "quote and cite" rule could
mistake a `web:` result for a curated primary source.

Add `web:` to the enumerated prefix list, mark it as an open web page
rather than a curated artifact, and note that the way to keep only
curated results is to pass `types=["doc","issue","pull_request","readme"]`
(`types=["web"]` is rejected with a 400).

Fixes firecrawl#213

@cubic-dev-ai cubic-dev-ai Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All reported issues were addressed across 1 file

Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic

Comment thread skills/firecrawl-developer-index/SKILL.md
Address cubic review on firecrawl#218: the earlier line 17 change said passing
`types=["doc","issue","pull_request","readme"]` "keeps only curated
results", but the Filters section right below still described the same
enumeration as the default — the workaround read as redundant.

Update the Filters `types` bullet to state plainly that when `types`
is unset the response also carries `web:` results, and that passing
an explicit list is the only way to exclude them. Keeps the section
consistent with the new line 17 note and with the reported behaviour
in firecrawl#213.
@gaurav-init

Copy link
Copy Markdown
Author

Thanks @cubic-dev-ai — you're right, the workaround read as redundant against the existing 'Defaults to all four' line. Pushed c8e383d which updates the Filters types bullet to say plainly that an unset types also returns web: results and that explicit enumeration is the only way to exclude them. Both lines are now consistent.

@cubic-dev-ai

cubic-dev-ai Bot commented Sep 5, 2026

Copy link
Copy Markdown

Thanks @cubic-dev-ai — you're right, the workaround read as redundant against the existing 'Defaults to all four' line. Pushed c8e383d which updates the Filters types bullet to say plainly that an unset types also returns web: results and that explicit enumeration is the only way to exclude them. Both lines are now consistent.

@gaurav-init Looks right. The Filters bullet now says plainly that an unset types also returns web: results and that only an explicit enumeration excludes them, which no longer reads as redundant against the line-17 workaround. The two statements line up: line 17 tells the agent what to pass, the types bullet explains why unsetting it doesn't do the job.

One tiny nit while we're here, only if you want it: the Filters bullet still starts with "which of doc, issue, pull_request, readme to search," which mirrors the default-enumeration phrasing from before. Since that sentence now sits right next to the "when unset it also returns web:" clause, a reader could briefly think the enumerated list is what you get by default before the next clause corrects it. Not worth a reword on its own — the meaning is unambiguous now — but if you ever touch this paragraph again, "which of the four curated kinds (doc, …) to search" removes that flicker entirely.

Both changes shipped in c8e383d are consistent and resolve the redundancy I flagged. No further action needed.

Tip: get faster answers by chatting with cubic’s review copilot in the review UI

@cubic-dev-ai cubic-dev-ai Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

0 issues found across 1 file (changes from recent commits).

Confidence score: 5/5

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

Shadow auto-approve: would auto-approve. Docs-only update to the developer-search skill: adds the web: id prefix and the curated-types workaround, with no code or behavior changes. The change is bounded and clearly beneficial.

Re-trigger cubic

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Developer Index skill documents four id prefixes; a fifth (web:) is the majority of unscoped results

1 participant