Repository navigation
chore: bump @anthropic-ai/claude-agent-sdk 0.3.146 -> 0.3.169 - #672
Merged
Merged
Conversation
Newest version within the repo's minimumReleaseAge (7-day) guard. Build + 893 tests green, no source changes needed. Notable in-range: 0.3.160 fixes abort-during-PostToolUse-hook hangs (relevant to our YARA hooks + abort); 0.3.162 folds Grep/Glob into Bash unless named in allowedTools (we're safe — BASE_ALLOWED_TOOLS lists both, keeping the search-scan hook alive). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
🧙 Wizard CIRun the Wizard CI and test your changes against wizard-workbench example apps by replying with a GitHub comment using one of the following commands: Test all apps:
Test all apps in a directory:
Test an individual app:
Show more apps
Results will be posted here when complete. |
gewenyu99
marked this pull request as ready for review
June 16, 2026 03:16
edwinyjlim
approved these changes
Jun 16, 2026
sarahxsanders
added a commit
that referenced
this pull request
Jun 16, 2026
Reconcile main's analytics instrumentation (#661, #666) and SDK bump (#672) against the branch's vitest migration: - analytics.test.ts: port main's live IS_PRODUCTION_BUILD getter mock and run_id/$session_id uuid sequencing to vitest (vi.hoisted + importOriginal, vi.fn, `as Mock`) - pnpm-lock.yaml: regenerate from merged package.json (claude-agent-sdk 0.3.169) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

Bumps
@anthropic-ai/claude-agent-sdk0.3.146 → 0.3.169.Not the absolute latest (
0.3.178): the repo'sminimumReleaseAge: 10080(7 days) guard blocks anything newer than ~Jun 8, so0.3.169is the newest eligible version.0.3.170–0.3.178come in the next bump once they age in.Verification
pnpm build+ full suite green (59 suites, 893 tests). No source changes required — the wizard already absorbs every behavior change in range.Changelog grooming (0.3.147 → 0.3.169)
Nothing to fix. One free win, one near-miss the wizard is already inoculated against.
0.3.160 — abort during a PostToolUse hook no longer hangs. Directly relevant: the wizard runs YARA scans as PostToolUse hooks (
yara-hooks.ts) alongside anabortController. Previously an abort mid-hook could hang the process; now it ends the turn with a finalresult. Straight win.0.3.162 — native builds fold
Grep/Globinto Bash unless named inallowedTools. This is the one that could have silently killed our security scanning — the YARA PostToolUse hook keys on theGreptool, which would never fire if searches routed through Bash. We're safe becauseBASE_ALLOWED_TOOLS(agent-interface.ts:1175) explicitly listsGrepandGlob, which keeps the dedicated tools registered. Guard rail: those two must stay inallowedToolsor the search-scan hook goes dark.0.3.149 —
options.envdocumented as replacing (not merging) the subprocess env, andCLAUDE_AGENT_SDK_VERSIONno longer dropped. We already spread...process.envintooptions.env(agent-interface.ts:867), so no change; we just get correct User-Agent/telemetry back.0.3.150 —
api_retryreportserror: 'overloaded'for 529 (was'rate_limit'). No impact — the wizard detects API errors by scanning assistant text, not by inspectingapi_retrymessages.Additive / low-impact, no action: 0.3.154 (stdio MCP no longer restarted every reconcile), 0.3.166 (MCP resource tools for runtime-added servers), 0.3.152 (
SessionStartreloadSkills+MessageDisplayhook), 0.3.163 (Stop/SubagentStopadditionalContext), 0.3.162 (refusals now carrystop_reason: 'refusal'— a future upgrade for our refusal detection if we want it). The rest are "parity with Claude Code" internal bumps.Deferred to the next bump (0.3.170–0.3.178, too fresh for the age guard)
disallowedToolsspecs (mcp__server,mcp__server__*) now honored instead of silently ignored. Once in, we could revisit thecanUseToolhard-gate workaround inagent-interface.ts:375-381.resultno longer dropped while a subagent runs; background/MCP task state restored on resume. Relevant to our subagent + orchestrator flows.claude-fable-5model +fablealias added to SDK model types.