Best possible solution:

Land after maintainer review accepts the current config-level proof or adds one credentialed small-model gateway profile run, while keeping the docs aligned with the harness behavior.

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

Not applicable; this PR adds a live-test harness mode rather than fixing a reproduced product bug. Source inspection confirms current main lacks the gateway small selector and the PR supplies config-level verification for the new path.

Is this the best way to solve the issue?

Yes, this is a reasonable owner-boundary fix: it reuses the existing direct small-model live helpers instead of creating a second model matrix. The remaining question is proof depth, not a clearly better implementation location.

AGENTS.md: found and applied where relevant.

Codex review notes: model gpt-5.5, reasoning high; reviewed against 5b84ebfc56c9.

Label justifications:

• P3: This is a low-risk docs and live-test harness improvement with no production runtime behavior change.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🦞 diamond lobster.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: The external-contributor proof gate does not apply to this MEMBER maintainer-labeled PR; the body still records terminal-style config/test proof and Crabbox changed-gate proof, with credentialed provider execution left untested.

PR surface:

Tests +165, Docs +4. Total +169 across 2 files.

What I checked:

• PR branch implements small gateway mode: The PR head parses OPENCLAW_LIVE_GATEWAY_MODELS=small, threads useSmall through provider scoping, dynamic small-ref loading, wanted-model filtering, capped selection, and not-found skip behavior. (src/gateway/gateway-models.profiles.live.test.ts:3540, 353a4444b695)
• PR branch adds focused harness tests: The PR head covers the small default cap, curated small-provider list, provider-filter intersection, filtered dynamic refs, and avoiding counts for small models outside a provider-scoped sweep. (src/gateway/gateway-models.profiles.live.test.ts:983, 353a4444b695)
• Docs describe the new recipe: The PR head documents the new gateway selector next to the existing direct small-model recipe and updates the cap wording for modern/all and small gateway sweeps. Public docs: docs/help/testing-live.md. (docs/help/testing-live.md:113, 353a4444b695)
• Current main does not already include the gateway small selector: Current main has the gateway model selector and the direct small-model recipe, but no OPENCLAW_LIVE_GATEWAY_MODELS=small docs or gateway useSmall path, so the PR is not obsolete on main. Public docs: docs/help/testing-live.md. (docs/help/testing-live.md:113, 5b84ebfc56c9)
• Existing small-model contract is already shared by direct live tests: Current main already defines the curated small priority refs, small cap, isSmallLiveModelRef, and selectSmallLiveItems; the direct live-model profile uses those helpers for OPENCLAW_LIVE_MODELS=small. (src/agents/live-model-filter.ts:39, 5b84ebfc56c9)
• Real behavior proof remains config-level: The PR body reports targeted live-config tests and an AWS Crabbox changed gate, while explicitly saying no credentialed small-model provider execution was run. (353a4444b695)

Likely related people:

• joshavant: Blame and git show --stat tie the current gateway live profile file, testing-live docs, and small-model filter helper surface to fbaa5a6f0a...; the checkout is grafted there, so this is strong for current-main routing but not full upstream provenance. (role: introduced current harness/helpers in local history; confidence: medium; commits: fbaa5a6f0a57; files: src/gateway/gateway-models.profiles.live.test.ts, src/agents/live-model-filter.ts, docs/help/testing-live.md)
• Vincent Koc: Current main includes recent live gateway maintenance by Vincent in 98f52dcc..., and this PR also extends that same live-profile harness area. (role: recent area contributor; confidence: high; commits: 98f52dcc00d2; files: src/gateway/gateway-models.profiles.live.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.