Follow-up: deferred review findings from PR #515

[philcunliffe]

Triage closure record for PR #515 (fix/issue-480, head d0c91a0372f3eedd81748f108db198e7a2842feb).

Review history

• Round 1 (e46ea60): 2 findings.
  • F1 (medium, regression-proofing on a privacy path): the demoted workspace
    key's own usableInBandCwd call site
    (hypaware-core/plugins-workspace/codex/src/exchange-projector.js:150) was
    unpinned; a relative workspace key could reach the shared .hypignore
    matcher, which resolves relative paths against the daemon's own cwd (Follow-up: deferred review finding from PR #466 (Codex live projector's in-band cwd reaches the .hypignore gate unpredicated) #471
    class defect on a new line). Fixed in-branch (a21b3ab):
    test/plugins/codex-rollout-cwd.test.js:272 (a relative workspace key is refused rather than resolved against the daemon (#480, #471)),
    mutation-verified.
  • F2 (low, doc accuracy): llp/0083-codex-live-cwd-from-rollout.decision.md
    still described the Follow-up: deferred review findings from PR #477 (ignored-workspace identity on newly-recorded rows; refusal warn fires every turn) #481 over-record limit as "clean in-band cwd,
    ignored declared workspace" only, when this PR widened the same case to
    the whole subscription route via the rollout. Fixed in-branch (a21b3ab).
• Round 2 (e3a169b): 0 findings. Both round-1 fixes re-verified with three
  targeted mutations, all reddening as expected; Codex workspace-cwd refusal is an ancestor test, not a byte test (#481, finding 2 of 2) #491's ancestor-test predicate
  confirmed untouched; LLP 0083/0160 cross-doc consistency confirmed.

Head d0c91a0 (post-round-2 state)

After round 2's clean pass, the branch merged origin/master again
(e3a169b..d0c91a0), pulling in PR #513 (10f4aa0, "Codex live projector
reads the cwd predicate from core, not a copy") among others. That merge was
never itself put through a review round. Triage independently verified it
introduces no new defect and does not regress either round-1 fix:

• git diff e3a169b d0c91a0 -- '*exchange-projector.js' shows only Codex live projector reads the cwd predicate from core, not a copy #513's own
  scoped change (usableInBandCwd now delegates its shape check to the shared
  sessionMetaCwd, exchange-projector.js:1286-1298); the ranking/gate logic
  this PR owns (exchange-projector.js:148-150, :889-909) is untouched.
• No unresolved merge-conflict markers anywhere in the tree.
• npm test on d0c91a0: 3267 pass / 0 fail / 1 skipped (better than
  round 2's reported baseline, since the merge also pulled in the
  leave-command fix, PR hyp leave: the daemon status query must answer, not throw #527/test/core/leave-command.test.js: 8 of 11 tests fail on master (exit 1 from hyp leave), red since the tests were introduced #512).
• npm run typecheck: clean.
• All @ref LLP … anchors touched by this PR's diff resolve
  (LLP 0083#workspace-key-ranks-last, LLP 0083#decision,
  LLP 0083#container-fallback-gap, LLP 0160#decision,
  LLP 0160#corrections-0083, LLP 0150#usable-cwd).
• llp/0083 and llp/0160 read consistently after the merge: the Codex live projector reads the cwd predicate from core, not a copy #513
  provenance correction and the Codex live projector: rank the workspace-key guess below the rollout cwd #515 ranking/limit edits both landed in the
  same bullets without contradiction.

Deferred items

None new. The one legitimate non-blocking backlog item visible in this
PR's review thread, that a workspace key outranked specifically by the
rollout (as opposed to an in-band cwd) is discarded silently rather than
logged (see llp/0083-codex-live-cwd-from-rollout.decision.md:~267,
"Three limits, stated rather than implied"), is already tracked as #492
and was explicitly judged acceptable, not a finding in round 1 (every
outcome remains observable on the row or in the drop log; only the
attribution is missing, and widening the warn would fire on every ordinary
subdirectory session per the frequency objection LLP 0160 settled). No new
issue is warranted for it.

The informational note in round 2 (the PR body's Docs section still names the
old PR #477 F4/D4 objection instead of LLP 0160, which now supersedes it) is
PR-description staleness, not a code or doc defect, and needs no tracking.

Backlink: #515

[philcunliffe]

Removing the neutral:fix label from this issue.

This is a closure record, not a deferred-findings backlog item: the triage of
PR #515 at head d0c91a0 found zero residual findings (both round-1 findings
were fixed in-branch and re-verified, and round 2 was clean). There is nothing
here to fix.

neutral:fix is the authorization signal for the issue-fix reconciler, so while
it was set, neutral issues reported this issue as needs-fix and the next tick
would have dispatched a fix agent against an empty work item. The label is
removed so the record stays readable without pulling work that does not exist.

The triage marker on PR #515 still points here, which reads correctly as
"triaged, nothing deferred". Close this whenever you like - neutral will not
reopen or act on it.