Best possible solution:

Keep current basic lean behavior for existing configs, and land the broader trim only behind an explicit strict opt-in profile or after maintainer-approved upgrade-break documentation and proof.

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

Yes for the review findings: current source and the PR diff show the existing three-tool lean contract and the PR's broadened deny set under the same flag. No live local-model run is needed to reproduce that compatibility change.

Is this the best way to solve the issue?

No. Broadening the existing boolean is a plausible mitigation, but the safer maintainable solution is an explicit strict opt-in profile or a clearly documented maintainer-approved compatibility break.

Full review comments:

• [P1] Preserve basic lean mode behind a strict opt-in — src/agents/local-model-lean.ts:6-30
  This expands the existing localModelLean boolean from the current three denied tools to a broad deny set that removes web, media, session/agent, gateway/node, image, and code-execution tools for every upgraded user who already enabled the flag. Keep the current basic behavior and put this larger trim behind an explicit strict profile, or document and approve the compatibility break before merge.
  Confidence: 0.91
• [P3] Include sessions_yield in the explicit docs list — docs/concepts/experimental-features.md:40
  The docs enumerate trimmed session/orchestration tools but omit sessions_yield, even though the new deny set removes it. If this explicit list remains, it needs to match the runtime contract so operators can predict which tools disappear.
  Confidence: 0.8

Overall correctness: patch is incorrect
Overall confidence: 0.9

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against f0f5da0e3941.

Label justifications:

• P2: This is a bounded agent/tool-surface change, but it affects existing local-model operator behavior and needs maintainer direction.
• merge-risk: 🚨 compatibility: The PR changes the semantics of an existing experimental config flag and can remove tools from upgraded users without a new opt-in.
• merge-risk: 🚨 message-delivery: The expanded deny set covers channel/session/orchestration tools, so existing lean-mode communication flows can lose direct tool access unless preserved carefully.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🐚 platinum hermit and patch quality is 🧂 unranked krab.
• status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Not applicable: The external-contributor proof gate does not apply to this maintainer-labeled member PR, though the body reports focused tests and AWS Crabbox changed-gate proof.

PR surface:

Source +24, Tests +55, Docs +7. Total +86 across 7 files.

What I checked:

• Current main lean contract: Current main still denies only browser, cron, and message for localModelLean, with Tool Search defaults layered separately. (src/agents/local-model-lean.ts:12, f0f5da0e3941)
• Current main delivery preservation: Current main passes localModelLeanPreserveToolNames into lean filtering so required runtime/delivery tools can survive the lean policy. (src/agents/agent-tools.ts:284, f0f5da0e3941)
• PR broadens the same boolean: The PR changes the existing deny set to 23 tool names under the same localModelLean boolean, including web, media, session/orchestration, gateway/node, image, and code-execution surfaces. (src/agents/local-model-lean.ts:6, e66d7c6713ea)
• Docs list mismatch: The PR runtime deny set removes sessions_yield, but the new explicit docs list omits it from the channel/session/orchestration tools it says lean mode trims. Public docs: docs/concepts/experimental-features.md. (docs/concepts/experimental-features.md:40, e66d7c6713ea)
• Strict-profile related PR is not canonical yet: The related strict-profile PR keeps basic lean behavior by adding an opt-in profile, but it is draft and currently conflicting, so it cannot safely supersede this PR. (10239eb74ee3)
• Feature history: Lean mode was introduced by merged PR history, then per-agent behavior and later Tool Search/default preservation were added in later area work. (src/agents/local-model-lean.ts, 13c97c5a8d04)

Likely related people:

• vincentkoc: Recent main history compacted lean local tool catalogs and the same person owns this PR and related strict-profile/media-trim lean work. (role: recent area contributor; confidence: high; commits: 8b03fd1f5f83, e66d7c6713ea; files: src/agents/local-model-lean.ts, src/agents/embedded-agent-runner/run/attempt.ts, docs/concepts/experimental-features.md)
• dutifulbob: Merged PR history for per-agent localModelLean added the current local-model lean module and tests. (role: introduced per-agent behavior; confidence: high; commits: 13c97c5a8d04; files: src/agents/local-model-lean.ts, src/agents/local-model-lean.test.ts, src/config/schema.help.ts)
• ImLukeF: Merged PR history introduced lean local model mode and the first three-tool lean contract. (role: original feature author; confidence: medium; commits: 4015138df93e; files: docs/gateway/local-models.md, src/config/schema.help.ts, src/config/types.agent-defaults.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.

Share on X: post this hatch
Copy: My PR egg hatched a 🥚 common Tiny Branchling in ClawSweeper.
Hatchability:

• Merged PRs are hatchable.
• Open PRs are hatchable when they are status: 👀 ready for maintainer look, status: 🚀 automerge armed, or labeled clawsweeper:automerge.
• Closed unmerged PRs are hatchable only when one of those hatchable labels is still present in the durable record.

About:

• Eggs appear after real-behavior proof passes. They are collectible flavor only.
• Review momentum changes the shell state: follow-up work warms it, re-review makes it wobble, and a clean final review lets it hatch.
• The hatch is seeded from this repository and PR number, so the same PR keeps the same creature; the reviewed head SHA can only change safe visual details.
• Rarity is just collectible sparkle: 🥚 common, 🌱 uncommon, 💎 rare, ✨ glimmer, and 🌈 legendary.