Public: anyone can read this; the crew can change it.
The duplicate fold never matches a sealed security report to another issue #14939
Check · Review · Land · Live done — Landed and live in production.
Sealed security reports are now left out of duplicate matching completely. The fold no longer closes a sealed report into another issue, and it never writes a "duplicate of #N" comment that would reveal one (#14908, from the sweep in #14768).
shipped production is serving it deploy 15342fb
Sealed security reports are now left out of duplicate matching completely. The fold no longer closes a sealed report into another issue, and it never writes a "duplicate of #N" comment that would reveal one (#14908, from the sweep in #14768).
What changed: duplicates() in src/crew/livetriage.ts now only reads issues with i.sealed = 0. I didn't match sealed reports against each other either. The person who filed a sealed report can read it but not someone else's sealed report. So a cross-reference between two sealed reports would still tell one reader that a report they can't see exists. As the issue's Risks note says, a sealed duplicate now stays open until a writer closes it by hand.
Test: a new case in test/livetriage.test.ts ("a sealed report is never matched, to an open issue or to another sealed one") has one open issue and two sealed reports with matching titles. After a triage pass all three are still open and none has a duplicate comment. The test fails without the fix in src/crew/livetriage.ts and passes with it.
Checks run: bun scripts/gate.ts --fix and bun x tsc --noEmit are clean. scripts/fast.ts named 390 files (about 34 minutes), so I picked the tests that cover this change instead: livetriage, dupes, sealed-issues, triage, trusted-fold and test-honesty. All 100 tests pass. The rest of the suite will run in the forge's check on this push.
To try it: bun test test/livetriage.test.ts -t "sealed report is never"
🤖 Generated with Claude Code
Sealed issues are filtered out of the duplicate fold at the query (i.sealed = 0), so no match and no "duplicate of #N" comment can involve one. The test covers an open issue plus two sealed look-alikes. Small, single-purpose, no new dependency, no page or terminal surface touched. I read the code and the diff only; I did not run the tests here. Next: (1) close sealed duplicates by hand with a visible hint for writers on the triage summary; (2) check other cross-reference writers (mentions, dependency links) for the same sealed leak.
Sealed security reports are now left out of duplicate matching completely. The fold no longer closes a sealed report into another issue, and it never writes a "duplicate of #N" comment that would reveal one (#14908, from the sweep in #14768).
What changed:
duplicates()insrc/crew/livetriage.tsnow only reads issues withi.sealed = 0. I didn't match sealed reports against each other either. The person who filed a sealed report can read it but not someone else's sealed report. So a cross-reference between two sealed reports would still tell one reader that a report they can't see exists. As the issue's Risks note says, a sealed duplicate now stays open until a writer closes it by hand.Test: a new case in
test/livetriage.test.ts("a sealed report is never matched, to an open issue or to another sealed one") has one open issue and two sealed reports with matching titles. After a triage pass all three are still open and none has a duplicate comment. The test fails without the fix insrc/crew/livetriage.tsand passes with it.Checks run:
bun scripts/gate.ts --fixandbun x tsc --noEmitare clean.scripts/fast.tsnamed 390 files (about 34 minutes), so I picked the tests that cover this change instead:livetriage,dupes,sealed-issues,triage,trusted-foldandtest-honesty. All 100 tests pass. The rest of the suite will run in the forge's check on this push.To try it:
bun test test/livetriage.test.ts -t "sealed report is never"🤖 Generated with Claude Code
Closes #14908