Conversation
Triages a change to what user data leaves the device — Sentry state masks, new MetaMetrics and Segment properties, error and breadcrumb strings that interpolate user values. Detection is the mask diff and is mechanical; the deliverable is what each newly unmasked field holds at runtime, since a bounded enum and an account address are indistinguishable to the mask. Its own domain rather than `security`, because the failure mode differs: security skills each have an adversary, and this one has none. The data is sent as designed, no boundary is crossed, and the defect is that nobody decided it should be. Named for what egresses. Privacy does not; data does.
MajorLift
added a commit
that referenced
this pull request
Sep 1, 2026
`data-egress-diligence` and its `CODEOWNERS` line move to #142. This PR is back to one domain: the three `security` skills that each have an adversary. Nothing else changes. The extracted skill cites no `knowledge/`, so its departure leaves no dangling reference, and the security domain's knowledge file stays with the two skills that cite it.
MajorLift
changed the title
privacy domain - /data-egress-diligence skill
`safe` was a verdict the skill leaves to the owners, while the bucket's own criterion is that a type bounds the value. `analytics-instrumentation` resolves to no skill since the analytics domain became `observability`.
…ction Sentry requires `consentDecisionMade && optedIn`. Under one shared label, a diff that drops the decision check but keeps `optedIn` read as the same protection removed and re-added, and printed no finding.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
Some user data leaves the device by design: a field unmasked in Sentry state, a new MetaMetrics or Segment event property, an error or breadcrumb string that interpolates a user value. When one of these is a defect, nothing announces it. The data is sent irrevocably, no test fails, no user reports it, and the mask list is not something a reviewer thinks to check.
The
securitydomain in #83 is organized around adversaries: someone ships malicious code, reaches a capability they should not, or constructs a URL the wallet did not expect. This failure mode has none. The data is sent as designed, no boundary is crossed, and the defect is that nobody decided it should be, so it gets its own domain.Overview
A
privacydomain with one skill,data-egress-diligence, which triages a change to what user data leaves the device: Sentry state masks inapp/scripts/constants/sentry-state.ts, new MetaMetrics and Segment event properties, and error and breadcrumb strings that interpolate user values.Detection is mechanical and is not the hard part. The mask diff says which fields became visible but not what they hold at runtime, and to a mask a bounded enum looks the same as an account address. So the skill sorts findings into bounded / needs-narrowing / must-not-egress with the evidence for each, and leaves the accept decision to the people who own it.
It is written to be discoverable rather than invoked, because nobody searches for a skill correcting a problem they do not know they have.
Showcase
No trial run yet, so nothing here speaks to its hit rate.
Ships
scripts/egress-delta.py, which diffs the outbound surface between two refs so the answer is computed rather than recalled.One
lint-skill-entrywarning on this skill,missing recommended section "## Workflow", shared with skills already onmain.