Best possible solution:

Compute or display an effective workspace path only where it is accurate for the selected mode/scope/backend, cover the output with focused command tests, and add redacted before/after CLI proof.

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

Yes. Current main and the latest release print only the configured workspaceRoot, while runtime layout uses the agent workspace for workspaceAccess: rw; I did not run a live Docker container in this read-only review.

Is this the best way to solve the issue?

No. The PR is a plausible narrow fix for rw, but it is not the best merge shape until the new effective field is accurate or conditional for every mode where it is displayed.

Full review comments:

• [P2] Compute effective roots for every displayed mode — src/commands/sandbox-explain.ts:199-202
  The new field is printed for all modes, but when workspaceAccess is not rw it falls back to sandboxCfg.workspaceRoot. Runtime does not use that parent as the effective workspace for normal scoped sandboxes; it resolves a scope key and then a per-session or per-agent sandbox directory. As written, the new effectiveWorkspaceRoot can still point users at the wrong filesystem location outside the rw case this PR fixes.
  Confidence: 0.86

Overall correctness: patch is incorrect
Overall confidence: 0.84

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 685c22dfb7e1.

Label changes:

• add P2: This is a focused CLI diagnostics bugfix for a sandbox boundary explanation with limited blast radius but real user confusion.
• add merge-risk: 🚨 other: Merging the PR as-is could add a new inaccurate sandbox filesystem diagnostic even though the runtime boundary itself is unchanged.
• add rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🦐 gold shrimp.
• add status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body has no real before/after CLI output, screenshot, recording, terminal transcript, or logs showing the changed sandbox explain behavior; contributor proof is needed before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Label justifications:

• P2: This is a focused CLI diagnostics bugfix for a sandbox boundary explanation with limited blast radius but real user confusion.
• merge-risk: 🚨 other: Merging the PR as-is could add a new inaccurate sandbox filesystem diagnostic even though the runtime boundary itself is unchanged.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🦐 gold shrimp.
• status: 📣 needs proof: The PR needs real behavior proof before ClawSweeper can clear the contributor ask. Needs real behavior proof before merge: The PR body has no real before/after CLI output, screenshot, recording, terminal transcript, or logs showing the changed sandbox explain behavior; contributor proof is needed before merge. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

PR surface:

Source +8. Total +8 across 1 file.

Security concerns:

• [low] Inaccurate sandbox boundary diagnostic — src/commands/sandbox-explain.ts:199
  The new effectiveWorkspaceRoot field can report the configured parent sandbox root instead of the actual scoped sandbox workspace, which can mislead users reasoning about sandbox filesystem exposure.
  Confidence: 0.82

What I checked:

• Repository policy read: Root AGENTS.md was read fully, and the scoped agents guide was read because the touched command imports agent/sandbox helpers; the review applied the policy requiring whole-path PR review and sandbox/security-boundary caution. (AGENTS.md:1, 685c22dfb7e1)
• Current main still has the reported mismatch: Current main builds payload.sandbox.workspaceRoot directly from sandboxCfg.workspaceRoot and prints that same field; there is no effective workspace field on main. (src/commands/sandbox-explain.ts:263, 685c22dfb7e1)
• Runtime effective workspace differs by mode and scope: Runtime layout resolves workspaceDir to the agent workspace only for workspaceAccess === "rw"; otherwise it uses a scoped sandbox workspace under the resolved sandbox root. (src/agents/sandbox/context.ts:90, 685c22dfb7e1)
• Scoped sandbox directory helper: Non-shared sandbox roots are resolved through resolveSandboxScopeKey and resolveSandboxWorkspaceDir, so the configured parent workspaceRoot is not the effective scoped workspace directory for normal session or agent scope. (src/agents/sandbox/shared.ts:25, 685c22dfb7e1)
• PR diff adds an always-visible effective field: The branch adds effectiveWorkspaceRoot and falls back to sandboxCfg.workspaceRoot whenever workspaceAccess is not rw, then prints that field in the human output. (src/commands/sandbox-explain.ts:199, cb9d59a4819b)
• Real behavior proof is absent: The PR body contains no before/after CLI output, screenshot, recording, or logs, and the live PR has failing Real behavior proof checks on the current head. (cb9d59a4819b)

Likely related people:

• steipete: Introduced sandbox explain in 9f9098406ccc95d3f67af0a758d6b5bec2448948, refactored sandbox workspace layout in 182afe9f594d22207aa452599ee41175747c1997, and current blame for the touched command resolves to a recent Peter-authored rewrite. (role: feature introducer and recent area contributor; confidence: high; commits: 9f9098406ccc, 182afe9f594d, 7613f6ebb3cb; files: src/commands/sandbox-explain.ts, src/agents/sandbox/context.ts)
• vignesh07: Added the shared bind-aware filesystem mount resolver that models effective workspace/agent/bind mounts and is the likely reusable seam for richer sandbox explain output. (role: adjacent owner; confidence: medium; commits: eafda6f52671; files: src/agents/sandbox/fs-paths.ts, src/agents/sandbox/workspace-mounts.ts)
• flashosophy: A recent merged PR rewrote the current tree shape for the touched command and sandbox helpers, but the sandbox explain behavior predates that unrelated TUI change. (role: recent file mover; confidence: low; commits: 3f1dea78753b; files: src/commands/sandbox-explain.ts, src/agents/sandbox/context.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.