Best possible solution:

Keep the narrow one-pass classification shape, but merge only after redacted terminal output or logs show an after-fix cached descriptor-backed plugin tool resolving and executing in a real setup.

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

Not applicable: this is a performance cleanup rather than a bug report. Source inspection confirms current main does two candidate-list filters, while PR head classifies the same candidates in one pass.

Is this the best way to solve the issue?

Yes for the code shape: one-pass classification is the narrowest maintainable change and preserves the existing execution order. Merge readiness still depends on contributor-supplied real behavior proof.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against a0e9ca1e9544.

Label justifications:

• P3: This is a small plugin-runtime performance cleanup with no new user-facing surface or urgent regression signal.
• rating: 🧂 unranked krab: Overall readiness is 🧂 unranked krab; proof is 🧂 unranked krab and patch quality is 🐚 platinum hermit.
• 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 lists focused Vitest and autoreview only; it needs redacted terminal output, logs, copied live output, or similar proof from a real cached descriptor-backed tool execution 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 +5. Total +5 across 1 file.

What I checked:

• Repository policy read: Read root policy and the scoped plugin policy; plugin cache/laziness guidance and the external-PR proof gate both apply to this review. (AGENTS.md:1, a0e9ca1e9544)
• Current main behavior: Current main still builds matchingNamedCandidates and unnamedCandidates with two separate filters before iterating named candidates then unnamed fallback candidates. (src/plugins/tools.ts:727, a0e9ca1e9544)
• PR head behavior: PR head classifies the same candidate groups in one loop and keeps the existing named-first, unnamed-fallback execution loop. (src/plugins/tools.ts:710, 5de49e3692d6)
• Adjacent cached descriptor tests: Existing tests cover cached descriptor execution, retained runtime registries, implicit unnamed registrations, shared declared names, and avoiding unrelated named factories before unnamed fallback. (src/plugins/tools.optional.test.ts:2025, a0e9ca1e9544)
• Live PR proof state: The PR body lists focused Vitest and autoreview only; no PR comments, reviews, labels, or body text provide redacted real cached descriptor execution output or a proof override. (5de49e3692d6)
• Related history: GitHub file history shows the cached descriptor resolver was recently shaped by merged PRs for dispatch-by-runtime-name and retained registries, with adjacent open cached-descriptor hardening PRs on the same files. (src/plugins/tools.ts:675, 3adbbe7c34b5)

Likely related people:

• zanni098: Authored the merged cached descriptor-backed dispatch fix that established the named-first then unnamed-fallback runtime-name matching invariant this PR preserves. (role: adjacent cached-tool fix contributor; confidence: high; commits: 3adbbe7c34b5; files: src/plugins/tools.ts, src/plugins/tools.optional.test.ts)
• luoyanglang: Authored the merged retained-registry fix in the same cached descriptor execution path and adjacent tests shortly before this PR. (role: recent area contributor; confidence: high; commits: 374076b5a8c5; files: src/plugins/tools.ts, src/plugins/tools.optional.test.ts, src/plugins/runtime/load-context.ts)
• vincentkoc: Co-authored and reviewed recent plugin descriptor/cache work and owns an overlapping open cached descriptor hardening PR touching the same resolver and tests. (role: adjacent area contributor; confidence: high; commits: 570e2db25210, 38b1a7c300e8; files: src/plugins/tools.ts, src/plugins/tools.optional.test.ts, src/plugins/tool-descriptor-cache.ts)
• steipete: Recent file history shows repeated work on plugin tool wrapping, descriptor metadata, and descriptor-cache-related helper surfaces around this runtime path. (role: feature-history contributor; confidence: medium; commits: f62a22ce560e, 643633c1e5fc, 4d54d196c99a; files: src/plugins/tools.ts, src/plugins/tool-descriptor-cache.ts, docs/plugins/sdk-overview.md)

• 🦀 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.