spidey-sense
← Incidents

Nimbus CLI (harness)

scrape-verse-delta.vercel.app · c_REPLACE_HARNESS · 20260821T090000Z-harness-changelog

resolved
BrokenHealthyresolved in 13m 4s

Contract score

Detected
0.80
12 rows
Heal preview
1.00
12 rows
Live canary
1.00
12 rows

Verification gate

PROMOTEDscore 0.80 → 1.00

Approved: every contract field is back within threshold (score 0.80 → 1.00).

Fill rate per field — before the heal, and in the heal’s preview. The tick is each field’s contract threshold. This is the evidence the gate scored.

versionrequired · nonEmptythreshold 95%
before
100%
after
100%
daterequired · dateishthreshold 95%
before
0%
after
100%
titlerequired · nonEmptythreshold 95%
before
100%
after
100%
permalinkrequired · urlthreshold 95%
before
100%
after
100%
tagoptional · nonEmptythreshold 90%
before
100%
after
100%
bodyoptional · nonEmptythreshold 80%
before
100%
after
100%

Heal prompt · 564 / 1000 chars

These fields stopped extracting correctly: date (valid in 0/12 rows; got "" (empty string), "" (empty string), "" (empty string)). Fix the extraction for those fields only. Keep every other field, and the output schema, exactly as they are. Still correct: version, title, permalink, tag, body — so the page loads and the row loop still matches. Only these fields moved. Expected values: date = an ISO date like "2026-08-18". Last selectors known to work: date was ".release time.release__date". Verify against: https://spidey-sense.vercel.app/harness/changelog?v=1
targeted:symptomguardrailworkingexpectedselectorsurl

Scraper Studio · scraper heal

Re-pointed the date selector at time[datetime] and now reads the attribute instead of the inline text node the redesign removed. Every other selector is unchanged.

Timeline

  1. 01

    Detected

    spidey-sense · contractAug 21, 09:00 UTC

    Broken: Required field `date` valid in 0/12 rows (0%), threshold 95%.

    evidence
    date: 0/12 valid (needs 95%) — got "" (empty string), "" (empty string), "" (empty string)
  2. 02

    Diagnosed

    spidey-sense · promptAug 21, 09:00 UTC

    Compiled a 564-char heal prompt targeting date.

    evidence
    These fields stopped extracting correctly: date (valid in 0/12 rows; got "" (empty string), "" (empty string), "" (empty string)). Fix the extraction for those fields only. Keep every other field, and the output schema, exactly as they are. Still correct: version, title, permalink, tag, body — so the page loads and the row loop still matches. Only these fields moved. Expected values: date = an ISO date like "2026-08-18". Last selectors known to work: date was ".release time.release__date". Verify against: https://spidey-sense.vercel.app/harness/changelog?v=1
  3. 03

    Healed

    Bright Data · scraper healAug 21, 09:11 UTC

    Scraper Studio proposed a rewrite and is awaiting approval.

    evidence
    Re-pointed the date selector at time[datetime] and now reads the attribute instead of the inline text node the redesign removed. Every other selector is unchanged.
  4. 04

    Verified

    spidey-sense · gateAug 21, 09:11 UTC

    Scored the proposed rewrite against the same contract: healthy, score 1.00.

  5. 05

    Promoted

    gate → scraper approveAug 21, 09:11 UTC

    Approved: every contract field is back within threshold (score 0.80 → 1.00).

  6. 06

    Confirmed

    spidey-sense · canaryAug 21, 09:13 UTC

    Re-ran the canaries: 12 rows, every contract field back within threshold.