docs(domain-skills/ebay): detect edge blocks, international pricing, fuzzy matches#362
Open
Lhy099 wants to merge 1 commit into
Open
Conversation
…fuzzy matches
The skill claimed "Chrome is NOT required" and the extractor assumed USD,
but field-testing from a CN/HK IP exposed three gaps that silently corrupt
agent output:
- Akamai edge returns HTTP 403/503 for low-reputation IPs (VPN, datacenter,
non-US residential) before any eBay HTML is served. The old text-only
`is_blocked()` never runs because `http_get` raises `HTTPError` first;
the agent sees a traceback, not a clean block signal.
- `s-card__price` ships raw localized strings ("HKD 297.54", "EUR 49.99"),
not just "$74.99". The old `\$([0-9,\.]+)` regex returned None for every
non-US-IP price.
- A query with no exact matches still renders ~60 fuzzy/recommended cards
with no structural difference. The only signal is a "No exact matches
found" hint; without checking it, agents silently treat substitutes as
successful matches.
Changes:
- Document `BROWSER_USE_API_KEY` + Chrome fallback as the way to bypass
the edge layer when the local IP is untrusted.
- Extend `is_blocked()` to accept HTTPError(403/429/503); add `safe_get()`
wrapper that returns None on either block layer.
- Rewrite price/strikethrough extraction to capture the raw inner text of
`s-card__price` instead of a `$`-prefixed numeric.
- Add `has_exact_match()` + a new section so agents surface fuzzy fallback
results to the user instead of silently substituting.
Field-tested 2026-05-15: query "dj posket4" from a HK IP → 0 exact matches,
60 fuzzy cards (5 with "posket" in title), all prices reported as HKD.
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The eBay scraping skill claimed "Chrome is NOT required" and the extractor assumed USD, but field-testing from a CN/HK IP exposed three gaps that silently corrupt agent output. This PR documents the gaps and patches the extractor.
Field-tested gaps
1. Akamai edge block (HTTP 403/503) is upstream of "Pardon Our Interruption"
Low-reputation IPs (VPN, datacenter, non-US residential) get rejected by Akamai before any eBay HTML is served. The old
is_blocked(html)only inspects text, buthttp_getraisesHTTPErrorfirst and never gets to call it — the agent sees a traceback instead of a clean block signal.2.
s-card__pricereturns localized strings, not just$xx.xxA US IP returns
"$74.99", a CN/HK IP returns"HKD 297.54", an EU IP returns"EUR 49.99". The old\$([0-9,\.]+)regex returnedNonefor every non-US-IP price.3. "No exact matches" is structurally invisible
A query with no exact match still renders ~60 fuzzy/recommended cards. The only signal is a small "No exact matches found" hint in the HTML. Without checking it, agents silently treat substitutes as successful matches.
Changes
BROWSER_USE_API_KEYfor fetch-use routing, real-Chrome fallback when the local IP is untrusted.is_blocked()to accepturllib.error.HTTPError(403/429/503); addsafe_get()wrapper that returnsNoneon either block layer.s-card__priceinstead of a$-prefixed numeric.has_exact_match()+ new section so agents surface fuzzy fallback results to the user instead of silently substituting.Field-test evidence (2026-05-15)
Query
dj posket4from a HK IP via harness Chrome path (becausehttp_get403d at the edge):posketin the title (Q posket); the other 55 were unrelatedHKD 297.54,HKD 273.98,HKD 446.32,HKD 422.75,HKD 297.54$74.99etc.) is preserved in the example output — both currency formats now parse correctly with the same extractor.Diff
Single file:
agent-workspace/domain-skills/ebay/scraping.md(+88 / -20).Summary by cubic
Docs update to the eBay scraping skill to handle Akamai edge blocks, localized currency prices, and “No exact matches” pages to prevent bad results. Adds
BROWSER_USE_API_KEY/Chrome fallback guidance, asafe_get()+ improvedis_blocked(), currency‑preserving price extraction, andhas_exact_match()to flag fuzzy results.Written for commit 77372b7. Summary will update on new commits.