Best possible solution:

Land the package-boundary cleanup only after maintainers explicitly accept the helper-export removal tradeoff; if real consumers exist, add a named deprecation or compatibility plan instead.

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

Yes. Source inspection shows current main exports repo-only Plugin SDK helper subpaths while those helper surfaces reach Vitest-backed test utilities; the PR body also reports the clean-package import crash that this fixes.

Is this the best way to solve the issue?

Yes, if maintainers accept the compatibility break. The patch uses the existing privateLocalOnly/public entrypoint split instead of adding another shim, but the exported-subpath removal is a maintainer compatibility decision.

AGENTS.md: found and applied where relevant.

Codex review notes: model gpt-5.5, reasoning high; reviewed against 798691779bee.

Label justifications:

• P2: This is a bounded package-boundary bugfix with meaningful Plugin SDK compatibility impact but no evidence of a live core outage.
• merge-risk: 🚨 compatibility: The diff intentionally removes public package export specifiers, which can break existing imports during upgrade.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (live_output): The PR body includes exact real behavior proof fields and reports build, release-check, npm pack measurement, and an import smoke over 303 Plugin SDK exports with 0 leaked helper exports; I did not rerun commands in this read-only review.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body includes exact real behavior proof fields and reports build, release-check, npm pack measurement, and an import smoke over 303 Plugin SDK exports with 0 leaked helper exports; I did not rerun commands in this read-only review.

PR surface:

Source +2, Tests +10, Docs +5, Config -22, Other -1. Total -6 across 11 files.

What I checked:

• Current main still exports the affected helper subpaths: Current main package.json still exposes Vitest-backed helper subpaths such as ./plugin-sdk/agent-runtime-test-contracts, ./plugin-sdk/plugin-test-contracts, ./plugin-sdk/provider-test-contracts, and ./plugin-sdk/testing as package exports. (package.json:573, 798691779bee)
• PR removes public exports and excludes helper artifacts: At the PR head, package.json adds file exclusions for the helper JS/d.ts outputs and the export block no longer includes the removed helper subpaths. (package.json:35, a907668ba889)
• PR uses the existing private/public SDK entrypoint split: The PR marks the helper subpaths as private local-only and changes listPluginSdkDistArtifacts to iterate publicPluginSdkEntrypoints, aligning package artifacts with public exports. (src/plugin-sdk/entrypoints.ts:111, a907668ba889)
• Guardrail coverage was added: The PR adds a focused contract guard that plugin-test-contracts, provider-test-contracts, and testing stay in the private local-only SDK entrypoint set. (src/plugins/contracts/plugin-sdk-package-contract-guardrails.test.ts:695, a907668ba889)
• Source inspection supports the reported Vitest dependency failure: Current helper surfaces reach repo-only Vitest-backed utilities while current package.json exports those helper subpaths, making the reported clean-package missing-vitest failure source-reproducible. (src/plugin-sdk/plugin-test-runtime.ts:52, 798691779bee)
• Compatibility policy applies: The root and scoped Plugin SDK policy treat plugin SDK public API changes and breaking removals as compatibility-sensitive, so this package-export removal needs explicit maintainer review even with green tests. (src/plugin-sdk/AGENTS.md:1, 798691779bee)

Likely related people:

• Steady-ai: Current-main blame and git log tie the package.json helper exports, SDK entrypoint helper generation, and package guardrail files to commit eb8f9b4. (role: current package-surface introducer; confidence: high; commits: eb8f9b46da39; files: package.json, src/plugin-sdk/entrypoints.ts, scripts/lib/plugin-sdk-entries.mjs)
• Vincent Koc: Current main already contains the same sessions-send-tool error-cause change at HEAD, so the PR's small agents hunk is adjacent to his recent fallback-error preservation work rather than the central SDK export cleanup. (role: recent adjacent owner; confidence: medium; commits: 798691779bee; files: src/agents/tools/sessions-send-tool.ts)
• Andy Ye: Blame shows the surrounding sessions_send fallback branch came from commit 81e7e8e, which explains the context for the small overlapping agents hunk. (role: introduced adjacent fallback path; confidence: medium; commits: 81e7e8ef24de; files: src/agents/tools/sessions-send-tool.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.

• Eggs appear after the PR passes real-behavior proof. It is here for vibes, not verdicts: it does not change labels, ratings, merge decisions, or automation.
• The shell reacts to review momentum: open follow-up work warms it up, re-review makes it wobble, and a clean final review lets it hatch.
• Hatchability usually comes from sufficient real-behavior proof, no blocking P0/P1/P2 findings, no security attention needed, and clean correctness. A merged PR is already final, so merge makes the egg hatchable independently.
• 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.