Build(deps): Update claude-agent-sdk requirement from <0.3,>=0.2.136 to >=0.2.157,<0.3

[dependabot]

Updates the requirements on claude-agent-sdk to permit the latest version.

Changelog

Sourced from claude-agent-sdk's changelog.

Changelog

0.2.159

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.281
• Pinned default model for e2e tests to claude-opus-5 to work around CI failures with the CLI's new default model (#1287)

0.2.158

New Features

• verbatim_prompts option: Added ClaudeAgentOptions.verbatim_prompts (default False). When True, user messages are delivered to the CLI exactly as written — no @path file expansion and no slash-command dispatch. This prevents untrusted text inlined into prompts from triggering unintended file reads or command execution. Works with query(), ClaudeSDKClient.connect(), and ClaudeSDKClient.query() for both string and async-iterable prompts. Requires CLI 2.1.248+; a warning is logged on older CLIs. (#1269)

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.280
• CI improvements: recompressed wheels and raised the PyPI pre-flight threshold (#1283)

0.2.156

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.276

0.2.155

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.275

0.2.154

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.274

0.2.153

New Features

• snapshot option for system prompts: Added a snapshot field to SystemPromptPreset and a new SystemPromptCustom typed dict. When snapshot is True, the session keeps the system prompt recorded on its first request, improving prompt-caching behavior across resumed sessions. When False, the prompt is rebuilt on every request, useful for iterating on append text. Requires CLI 2.1.257+ (#1268)

Internal/Other Changes

• Updated bundled Claude CLI to version 2.1.273

0.2.152

... (truncated)

Commits

• See full diff in compare view

[pengfei-threemoonslab]

Review at 45b7bc1eff86230cdd114f807bb35181a8d611b4: the one-line SDK floor change is scoped to the developer adoption harness. No code defect found. All 169 harness tests pass locally, and fresh Agents Shipgate verification returns control_state=complete.

Validation remains incomplete: harness/requirements.txt explicitly requires smoke plus a paid Sonnet cell against the new SDK before raising this floor. The existing mocked tests do not demonstrate live streaming, tool events, and cost reporting with 0.2.157. No Anthropic credentials are available in this review session. The repository owner explicitly requested leaving this PR open pending live validation. No merge attempted.

[pengfei-threemoonslab]

Rechecked the updated revision f9fb6ebd834fed8ed2b66aa33dc6b7bac18914d3: the PR now raises the SDK floor to 0.2.157 and still changes only harness/requirements.txt. The adjacent requirement to rerun the smoke test and a paid Sonnet cell still applies. Static verification passes, but it cannot establish live SDK event/cost behavior.

Per the maintainer's explicit instruction, this PR remains open pending live validation. No paid run was attempted and no merge is planned without that evidence.

[dependabot]

Superseded by #885.