chore(acp): bump codebuddy, dimcode, grok and kilo registry pins to probed versions

[kaizhou-lab]

Summary

Scheduled ACP Registry version sync. Four npx pins drifted since #980, each backed by a fresh serial ACP probe of the exact pinned version. The diff is the lock file plus the one lock-derived test assertion that embeds codebuddy's version.

backend	package	old → new	initialize	session/new
codebuddy	@tencent-ai/codebuddy-code	2.147.0 → 2.148.0	ok (protocolVersion 1)	auth required (-32000, data.category: auth)
dimcode	dimcode	0.3.32 → 0.5.1	ok (agentInfo version 0.5.1)	auth required (-32000, "Provider credentials are required")
grok	@xai-official/grok	1.0.24 → 1.0.27	ok (protocolVersion 1)	auth required (-32000, "no auth method id provided")
kilo	@kilocode/cli	7.5.16 → 7.6.0	ok (agentInfo Kilo 7.6.0)	succeeded unauthenticated

All four meet the release-lock criterion: initialize succeeds, and session/new either succeeds or returns a clearly classified authentication requirement. Probes ran serially with no inherited HOME or credentials.

Two of these cross a minor series, so their entrypoints were exercised rather than assumed. dimcode jumps from 0.3.32 straight to 0.5.1, skipping the entire 0.4.x line — the Registry still declares the bare acp subcommand, and that exact invocation completed initialize with agentInfo self-reporting 0.5.1 before returning its usual credentials error. kilo moves 7.5.16 → 7.6.0 and its acp entrypoint still opens a session advertising model, thought_level, and mode. Neither session catalog is persisted; skill step 11 leaves those columns to the runtime.

dimcode continues to self-report agentInfo.title as "DimAgent" against a public listing of "DimCode" — a standing vendor quirk across this whole version jump, not an AionCore defect.

Snapshot structural diff (against v2026.09.09-cfa1ca8): the id set holds at 40 and every change is version churn — no distribution type added or removed, and args/env are byte-identical for every npx entry, including for the two minor-series jumps above. Non-lock movement is report-only: claude-agent-acp 0.76.0, codex-acp 1.11.0, factory-droid 0.216.0, qwen-code 0.23.3, plus new binary artifacts for devin, kilo's binary channel, and kimchi.

Derived assertion updated: registry_npx_lock.rs pins codebuddy's exact version inside a --package-form argument list, so it moves with the lock — 2.147.0 → 2.148.0. All four outgoing versions were scanned across crates/**/*.rs with fixed-string matching before staging; codebuddy's is the only real assertion and the other three have no hits. npx_cache_repair.rs keeps its own version literals: those are cache-path hash fixtures, not lock assertions.

The other 7 Registry-pinned packages (autohand, deepagents, dirac, glm-acp-agent, nova, pi, sigit) match the snapshot exactly. Drifted but not upgraded: none. mimo-code remains the one non-Registry builtin (no registry_json_id), excluded from drift reconciliation.

Standing watches (unchanged, no action in this PR):

• kimchi (listed 2026-09-09) stays deferred — binary-only, so it cannot enter the npx release lock; its binary artifact moved again this window. Integrating it needs a binary install/update path plus a seed migration, which is a human-reviewed change.
• sigit's npm scope rename (@smbcloud/sigit → @getsigit/sigit) is still only a vendor stderr notice; the Registry declares the old scope, so the pin is untouched.

Registry snapshot

• Audit pinned to release tag v2026.09.10-797d203 of agentclientprotocol/registry, fetched via the versioned CDN path for reproducibility.
• 40 ids in the raw snapshot: no newly listed and no delisted agents versus the previous run. (antigravity-acp and kimchi remain listed and deferred as binary-only; fast-agent and minion-code remain listed but uvx-only and therefore out of scope.)

Validation

• just migration-check — pass
• just lint-fix (cargo fix + clippy --fix --workspace -D warnings) — clean
• just fmt — clean
• Local cargo nextest intentionally skipped, by standing policy for lock-only bumps (established 2026-08-11). The Test check on this PR is the authority for this change: the merge decision depends on CI rather than the local run, and this host's load only manufactures timeout-shaped test failures, which nothing in the local steps above is subject to.

Logging

No logging changes: lock version bumps plus one test assertion; existing startup/session error paths already identify a failing agent by backend.