Best possible solution:

Merge the focused caller-scope registration change once exact-head checks finish, retaining one EXIT-owned cleanup list and the wrapped command's original exit status.

Do we have a high-confidence way to reproduce the issue?

Yes. On current main, tmp="$(mktempfile)" executes the helper in a subshell, so the parent TMPFILES array remains unaware of the generated path and cannot remove it on exit.

Is this the best way to solve the issue?

Yes. Assigning into a named caller variable fixes the root Bash scope issue, converts every affected caller, preserves the existing cleanup lifecycle, and follows the merged sibling installer pattern instead of adding a competing cleanup mechanism.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 67e1d4391153.

Label changes:

• add P2: The PR fixes bounded temporary-file leakage during failed installer operations without affecting core runtime availability or persisted user data.
• add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🐚 platinum hermit and patch quality is 🐚 platinum hermit.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: The PR is marked as maintainer work; its body nevertheless records Blacksmith Testbox and source-blind terminal validation of exit-code preservation and temporary-path removal.

Label justifications:

• P2: The PR fixes bounded temporary-file leakage during failed installer operations without affecting core runtime availability or persisted user data.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🐚 platinum hermit and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: The PR is marked as maintainer work; its body nevertheless records Blacksmith Testbox and source-blind terminal validation of exit-code preservation and temporary-path removal.

PR surface:

Tests +32, Other +2. Total +34 across 2 files.

What I checked:

• Current-main bug: mktempfile appends to TMPFILES, while each affected caller captures its output with command substitution; the helper therefore mutates a subshell copy rather than the parent array used by the EXIT trap. (scripts/install.sh:51, 67e1d4391153)
• Complete caller migration: The branch changes every mktempfile caller—remote scripts, spinner output, quiet-step logs, npm diagnostics, and three NodeSource branches—to use caller-scope assignment. (scripts/install.sh:51, 25af7f54deff)
• Regression coverage: The added test forces a downloaded script to exit 42, verifies the exit status is preserved, verifies the generated file is removed by the EXIT trap, and guards against reintroducing command-substitution calls. (test/scripts/install-sh.test.ts:58, 25af7f54deff)
• Sibling implementation precedent: Merged PR fix(install-cli): track mktemp paths and clean up on EXIT #103725 registers temporary paths directly in the parent shell for install-cli.sh, documenting the same command-substitution subshell invariant. (scripts/install-cli.sh:32, 95b205eac28d)
• No touched-file base drift: Current main has no changes to either touched file after the PR base, so the reviewed patch still targets the same source and test context. (scripts/install.sh:51, 67e1d4391153)
• Protected review routing: The live PR carries the maintainer label, which prevents conservative ClawSweeper auto-close even though the patch review is clean. (25af7f54deff)

Likely related people:

• steipete: Current-main blame and history tie the primary installer cleanup helper and affected callers to Peter Steinberger's imported installer work; he also authored this follow-up. (role: introduced current behavior; confidence: high; commits: 0120cacf9e4e; files: scripts/install.sh, test/scripts/install-sh.test.ts)
• SebTardif: Authored the merged sibling CLI-installer cleanup that established direct parent-shell temporary-path registration and failure-path coverage. (role: introduced sibling behavior; confidence: high; commits: 95b205eac28d; files: scripts/install-cli.sh, test/scripts/install-cli.test.ts)

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.