You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Preserves the full parse-address output and original address text when scanning broker profiles, then flattens the current street and zipCode so opt-out actions can reference them. The new field mappings use real extracted values when available and fall back to the existing generated values otherwise.
This is a proof of concept for making peoplesearchnow.com the PeopleFinders cluster's opt-out anchor. It includes a sanitised PeopleSearchNow-style HTML fixture which verifies that the current address is selected instead of address history, and that the street, city, state and ZIP survive extraction.
Native persistence and broker-rule schemas live outside this repo and still need verification.
Testing Steps
Run npm run test-unit -w injected.
Run npm run test-int -w injected -- --project windows --grep "PeopleSearchNow-style" --reporter list.
Run npm run tsc && npm run tsc-strict-core.
Run npm run build -w injected.
Checklist
Please tick all that apply:
I have tested this change locally
I have tested this change locally in all supported browsers
This change will be visible to users
I have added automated tests that cover this change
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Web Compatibility & Security Assessment
Risk Level: Low Risk — Broker-protection-only changes (extract/aggregate/fill-form) with no browser API overrides, messaging transport changes, or global prototype/DOM side effects; impact is confined to config-gated data-broker opt-out flows.
formatStreetAddress() reassembles components via parse-address normalization (e.g. street → St). Opt-out forms that validate against the exact page text may reject the filled value even when extraction succeeded. Consider falling back to the unparsed street line from fullAddress when broker configs require verbatim text.
Address deduplication still keys on `${city},${state}`. Two distinct full addresses in the same city/state will collapse to one entry in addresses[]. More visible now that entries carry streetAddress/fullAddress.
currentAddress is taken from pre-sort page order (addressFull[0] / addressFullList[0]) while addresses[] is sorted — intentional per unit tests, but broker configs must target the "current" anchor (XPath in PeopleSearchNow test) rather than relying on list position after aggregation.
All runtime files
—
info
No API shims, prototype mutations, observers, or viewport/meta changes. Broker protection is remote-config gated and only executes native-orchestrated actions on broker pages — no general third-party site compatibility surface.
Tests only
—
info
Integration + unit coverage added (45 specs pass locally). PeopleSearchNow XPath selector correctly scopes to "Current Address:" and excludes "Used to Live:" history.
Security Assessment
File
Lines
Severity
Finding
All runtime files
—
info
No changes to captured-globals.js, message bridge, origin validation, or nativeData handling. Extracted address data flows only through existing actionCompleted broker-protection messaging — no new exfiltration vectors.
New street / zipCode field types write page-scraped PII into broker opt-out form inputs — intentional and scoped to broker automation. Falls back to existing generators when data is absent (data.street || generateStreetAddress()).
...parsed spreads all parse-address output into result objects forwarded to native. Page-controlled DOM text could yield unexpected parser fields, but output stays within the broker-protection trust boundary and is not executed.
Pre-existing setValueForInput() uses uncaptured Object.getOwnPropertyDescriptor, window.HTMLInputElement, and Event — not introduced by this PR; acceptable in broker-protection context where page scripts are already active.
Recommendations
(warning) Add a unit test where two full addresses share city/state and confirm dedup behaviour is acceptable, or key the Map on fullAddress/streetAddress when present.
(warning) Document in broker config guidance that formatStreetAddress may normalize abbreviations; provide a config escape hatch (e.g. fill fullAddress street line verbatim) if brokers reject normalized values.
(info) Consider replaceAll(' ', marker) (or a global regex) in fullAddressStrings if multi-line addresses become common.
(info) Coordinate privacy-configuration broker action JSON to use the new street/zipCode fill types and PeopleSearchNow-style addressFull XPath anchor before shipping.
Verdict: Approve from a web-compat/security perspective. Changes are well-scoped, tested, and do not touch security-critical infrastructure. Residual risks are broker-specific extraction/fill fidelity, not general web breakage.
The reason will be displayed to describe this comment to others. Learn more.
warning (broker fill fidelity):formatStreetAddress() normalizes via parse-address (e.g. street → St). If an opt-out form validates against the exact scraped span text, the normalized value may be rejected. Consider a verbatim fallback from the street line of fullAddress when available.
The reason will be displayed to describe this comment to others. Learn more.
warning (data loss): Dedup key is still `${city},${state}`. With richer address objects, two distinct streets in the same city/state will collapse. Worth a test or switching the key to fullAddress/streetAddress when those fields exist.
The reason will be displayed to describe this comment to others. Learn more.
info: Only the first is protected before list splitting. Multi-line addresses with several breaks may still split incorrectly when a custom separator is configured.
The reason will be displayed to describe this comment to others. Learn more.
Stale comment
Injected PR Evaluation: Web Compatibility & Security
Re-assessed on synchronize (a9e792496). Latest commit updates an existing integration-test expectation to match the richer address profile — no runtime changes.
Web Compatibility Assessment
injected/src/features/broker-protection/extractors/address.js (149–152) — warning: formatStreetAddress() reassembles parsed components with normalized abbreviations (e.g. "street" → "St"). Opt-out forms that require verbatim page text may reject the filled value even when fullAddress preserves the original string.
injected/src/features/broker-protection/actions/extract.js (391) — warning: Address deduplication still keys on `${city},${state}`. With richer full-address objects, distinct streets in the same city/state collapse to one entry in addresses[].
injected/src/features/broker-protection/extractors/address.js (136–139) — info: fullAddressStrings() protects only the first before list splitting; additional newlines are collapsed via replace(/\s+/g, ' ') at parse time.
injected/src/features/broker-protection/actions/fill-form.js (86–89) — info: New street / zipCode field types fall back to generated values when data is missing (unlike cityState, which errors). Intentional for opt-out resilience, but a partial extraction could submit synthetic address data.
No general web-compat regressions — Changes are scoped to broker-protection extract/fill logic on data-broker pages. No browser API overrides, prototype mutation, observers, or page-wide DOM side effects.
Security Assessment
No security-critical path changes — No edits to captured-globals.js, wrapper utilities, message bridge, messaging transports, or origin validation.
injected/src/features/broker-protection/actions/extract.js (116–122, 410–416) — info: Extended scrape output (...parsed, top-level street/zipCode) flows through the existing SuccessResponse → native messaging path. Data is page-derived; no new exfiltration vectors beyond pre-existing broker-protection extraction.
injected/src/features/broker-protection/actions/fill-form.js (86–89) — info: Pre-existing setValueForInput uses uncaptured Object.getOwnPropertyDescriptor / Event (established autofill pattern). No new dynamic code execution, postMessage, or network requests.
Messaging hygiene — No object spreading of unknown payloads; no nativeData in outgoing messages.
Risk Level
Low Risk — Broker-protection-only changes extending address extraction and form-fill field types, with solid unit/integration coverage and no API-override or security-boundary modifications.
Recommendations
(warning) When configuring opt-out forms, prefer fullAddress (verbatim) over street/streetAddress if a broker validates against exact page text; document this in broker config guidance.
(warning) Consider dedup keying on fullAddress or streetAddress when available, not just city,state, to avoid losing distinct addresses in the same locale.
(info) Add a unit test for multi-line addresses with more than one if brokers emit that format.
(info) Consider explicit error (not generated fallback) for street/zipCode when extraction is required for a given broker step.
Tests: 45/45 broker-protection extract + fill-form unit specs pass locally.
Preserve extracted street and ZIP data so broker opt-out forms can use real addresses while retaining generated fallbacks.
Co-authored-by: Cursor <[email protected]>
Add a sanitised broker-style fixture to verify the current full address is selected and retained end to end.
Co-authored-by: Cursor <[email protected]>
Return a clean {streetAddress, city, state, zip, fullAddress} shape from
extractAddressFull instead of leaking all parse-address internals, and split
address lists by delimiter before collapsing intra-address whitespace (dropping
the newline-marker hack). Document the city/state dedup and the street/zipCode
generated fallback, and add an end-to-end opt-out form-fill integration test.
Fixture names follow the existing scenario-based convention (no broker name).
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
formatStreetAddress() rebuilds the street line from parse-address tokens, which normalizes abbreviations (e.g. "Street" → "St"). Brokers that validate opt-out forms against verbatim page text may reject the normalized value. fullAddress preserves the original string but is not used by fillForm today.
Address deduplication still keys on ${city},${state} only. Multiple distinct streets in the same city/state collapse to one entry in addresses. Documented as acceptable for city/state matching, but the surviving entry may not be the user's actual street. Form fill is unaffected because currentAddress is taken from page order separately.
street and zipCode fall back to generated values when extraction is empty (same pattern as existing $generated_*$ types). A generated street/zip that does not match the broker record could cause opt-out validation failure, but prevents leaving required fields blank.
No API overrides, prototype mutations, or global shims. Changes are confined to broker-protection extract/aggregate logic, which only runs when the feature is enabled and native schedules an action on a broker page.
New street/zipCode fields are explicitly constructed from scraped DOM text and returned to native via existing actionCompleted messaging. No nativeData forwarding; no new message schemas or bridge changes.
Page-derived address components are written into broker opt-out form fields — intentional, scoped behavior. Values pass through existing setValueForInput() (native setter + synthetic events); no new eval, postMessage, or network requests.
parse-address parses untrusted page text. Output is used only for broker opt-out automation, not for security decisions. Parsing failures are filtered by the parsed.city guard.
—
—
info
No changes to captured-globals.js, message-bridge trust boundaries, origin validation, load()-time hooks, or shouldExemptMethod(). Pre-existing uncaptured Object.getOwnPropertyDescriptor / new Event() usage in fill-form.js is unchanged.
Risk Level
Low Risk — Broker-protection–scoped extraction and form-fill changes only; no browser API overrides, messaging security changes, or cross-site impact. Blast radius is limited to configured broker opt-out flows.
Recommendations
(warning) Monitor PeopleSearchNow opt-out validation with normalized streetAddress values. If brokers reject abbreviated forms, consider a fullAddress-based fill type or a verbatim street-line extractor for the street field.
(warning) Document the city/state dedup limitation in broker config guidance so operators know addresses is a match aid, not a complete history list.
(info) If newline-separated address lists appear in the wild, add an explicit separator to the broker config or extend ADDRESS_LIST_SEPARATOR.
(info) 48 broker-protection unit specs pass locally (broker-protection-extract.js, broker-protection-fill-form.js).
The reason will be displayed to describe this comment to others. Learn more.
warning (web compat):formatStreetAddress() normalizes tokens via parse-address (e.g. "Street" → "St"). If a broker's opt-out form validates against verbatim page text, this normalized value may be rejected even though fullAddress preserves the original. Consider exposing a verbatim street fill path if PeopleSearchNow validation fails in production.
The reason will be displayed to describe this comment to others. Learn more.
warning (web compat): Dedup still keys on city,state only — distinct streets in the same city/state collapse to one addresses entry. Acceptable for city/state matching, but worth noting in broker config docs. Form fill is safe because currentAddress is sourced from page order above.
Add top-of-fixture comments noting the markup is a trimmed subset of a real
PeopleSearchNow page (matching the Veripages fixture convention), and replace
the real broker-record name with the standard Smith placeholder.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
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
semver-minorNew feature — triggers minor version bump
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.
github-actions
Bot
added
the
semver-minor
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
Asana Task/Github Issue: N/A
Description
Preserves the full
parse-addressoutput and original address text when scanning broker profiles, then flattens the currentstreetandzipCodeso opt-out actions can reference them. The new field mappings use real extracted values when available and fall back to the existing generated values otherwise.This is a proof of concept for making peoplesearchnow.com the PeopleFinders cluster's opt-out anchor. It includes a sanitised PeopleSearchNow-style HTML fixture which verifies that the current address is selected instead of address history, and that the street, city, state and ZIP survive extraction.
Native persistence and broker-rule schemas live outside this repo and still need verification.
Testing Steps
npm run test-unit -w injected.npm run test-int -w injected -- --project windows --grep "PeopleSearchNow-style" --reporter list.npm run tsc && npm run tsc-strict-core.npm run build -w injected.Checklist
Please tick all that apply: