Add Python low-level tool-definition E2E test [4/6]

[edburns]

Summary

Adds Python low-level tool-definition E2E coverage in python/e2e/test_tools_e2e.py as part of the PR-1692 breakout.

This PR is related to issue #1682 but does not fix #1682.

What changed

• Adds test_low_level_tool_definition coverage in Python E2E tests.
• Reuses the shared snapshot introduced by PR Add Java low-level tool definition E2E test and skill [1/6] #1721:
  • test/snapshots/tools/low_level_tool_definition.yaml
• Keeps low-level tool registration scoped to tools exercised by that snapshot (set_current_phase, search_items).
• Validates decoded keyword argument handling in the tool invocation path.

Dependency / sequencing

• Depends on PR Add Java low-level tool definition E2E test and skill [1/6] #1721 being merged first (already merged).
• Part of the Start iterating on #1682 On branch edburns/1682-java-tool-ergonomics #1692 breakout plan (PR 4 of 6).

Related

• Issue: Java: Ergonomics: Defining tools #1682 (not fixed by this PR)
• PR Add Java low-level tool definition E2E test and skill [1/6] #1721
• Original combined PR: Start iterating on #1682 On branch edburns/1682-java-tool-ergonomics #1692

[Copilot]

Pull request overview

Adds Python E2E coverage for the “low-level tool definition” scenario by introducing a new test in the Python E2E suite that reuses the shared replay-proxy snapshot under test/snapshots/tools/low_level_tool_definition.yaml.

Changes:

• Adds test_low_level_tool_definition to exercise explicit tool registration with two custom tools (set_current_phase, search_items).
• Configures a per-session tool allowlist via ToolSet().add_custom("*").add_builtin("web_fetch") to mirror the shared scenario setup.
• Asserts tool argument decoding/availability (keyword) and validates observable side effects (current_phase) after tool execution.

Show a summary per file

File	Description
python/e2e/test_tools_e2e.py	Adds a new E2E test that registers two tools and validates tool invocation + results using the shared tools/low_level_tool_definition snapshot.

Copilot's findings

• Files reviewed: 1/1 changed files
• Comments generated: 0

[github-actions]

Cross-SDK Consistency Review ✅

This PR is part of a well-coordinated 6-part breakout series (#1721–#1728) adding low_level_tool_definition E2E coverage across all six SDKs. Cross-SDK consistency is maintained by the series as a whole:

PR	SDK	Status
#1721	Java + shared snapshot	merged
#1724	Go [2/6]	open
#1725	Node.js [3/6]	open
#1726	Python [4/6]	this PR
#1727	Rust [5/6]	open
#1728	.NET [6/6]	open

Python implementation analysis

API naming follows Python conventions correctly, parallel to other SDKs:

• ToolSet().add_custom("*").add_builtin("web_fetch") matches Java's new ToolSet().addCustom("*").addBuiltIn("web_fetch")
• PermissionHandler.approve_all matches Java's PermissionHandler.APPROVE_ALL
• invocation.arguments matches Java's invocation.getArguments()

Tool scope intentionally limited to set_current_phase + search_items — the two tools actually exercised by the shared snapshot. This matches the stated design of PRs #1724, #1725, #1727, and #1728.

Minor observation (pre-existing, not introduced here)

The Java test (LowLevelToolDefinitionIT.java, merged in the series' first PR) includes an extra grepOverrideTool definition that the shared snapshot never exercises. The subsequent SDKs (#1724–#1728, including this one) deliberately omit it. This creates a minor divergence between Java and the rest, but is intentional and acknowledged in this PR's description. The Python test's more focused approach is actually the pattern adopted for the remainder of the series.

No action required on this PR. The change is self-consistent and correctly aligned with the cross-SDK rollout plan.

Generated by SDK Consistency Review Agent for issue #1726 · sonnet46 1.2M · ◷