Fix Python permission handler in custom agents guide

[patniko]

Fix

The Python on_permission_request lambda in the custom agents guide only accepted one argument (lambda req:), but the handler signature requires two positional arguments (request, invocation). This would cause a TypeError at runtime.

Before: "on_permission_request": lambda req: {"kind": "approved"}
After: "on_permission_request": lambda req, inv: {"kind": "approved"}

Review performed

Verified all code samples in the guide against actual SDK source across all 4 languages:

• ✅ Node.js/TypeScript — all API signatures, event types, and data fields accurate
• ✅ Python — fixed the permission handler bug; all other patterns correct
• ✅ Go — struct fields, pointer patterns, and event handling accurate
• ✅ .NET — typed event classes, pattern matching, and config properties accurate
• ✅ Event types table and configuration reference verified against generated types

[Copilot]

Pull request overview

Updates documentation and E2E coverage around permissions/session-resume/streaming behavior across SDKs, plus adds a new agentic workflow for cross-repo issue analysis.

Changes:

• Fix Python on_permission_request example signature in the custom agents guide.
• Add/adjust E2E tests + snapshots for streaming delta behavior (including after session resume) and for continued stateful conversation after resume.
• Standardize CI detection in harnesses to GITHUB_ACTIONS=="true" and add a cross-repo issue analysis workflow (+ compiled lock file).

Reviewed changes

Copilot reviewed 22 out of 22 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/snapshots/streaming_fidelity/should_produce_deltas_after_session_resume.yaml	Adds snapshot for streaming deltas after resuming a session
test/snapshots/session/should_resume_a_session_using_the_same_client.yaml	Extends resume snapshot with a follow-up message to validate statefulness
test/snapshots/session/should_resume_a_session_using_a_new_client.yaml	Extends resume snapshot (new client) with a follow-up message to validate statefulness
test/snapshots/session/should_receive_streaming_delta_events_when_streaming_is_enabled.yaml	Removes snapshot for deleted streaming-delta test in session suite
test/snapshots/session/should_pass_streaming_option_to_session_creation.yaml	Removes snapshot for deleted “streaming option accepted” test
test/harness/replayingCapiProxy.ts	Switches CI strictness check to GITHUB_ACTIONS
python/e2e/testharness/context.py	Switches fake-token CI detection to GITHUB_ACTIONS
python/e2e/test_streaming_fidelity.py	Adds Python E2E streaming fidelity test module
python/e2e/test_session.py	Adds stateful follow-up assertions after resume; removes older streaming tests
nodejs/test/e2e/streaming_fidelity.test.ts	Adds “deltas after session resume” coverage
nodejs/test/e2e/session.test.ts	Adds stateful follow-up assertions after resume; removes older streaming tests
nodejs/test/e2e/harness/sdkTestContext.ts	Introduces isCI based on GITHUB_ACTIONS and reuses it
go/internal/e2e/testharness/context.go	Switches fake-token CI detection to GITHUB_ACTIONS
go/internal/e2e/streaming_fidelity_test.go	Adds Go E2E streaming fidelity tests
go/internal/e2e/session_test.go	Adds stateful follow-up assertions after resume; removes older streaming tests
dotnet/test/StreamingFidelityTests.cs	Adds .NET E2E streaming fidelity tests
dotnet/test/SessionTests.cs	Adds stateful follow-up assertions after resume; removes older streaming tests
dotnet/test/Harness/E2ETestContext.cs	Switches CI detection to GITHUB_ACTIONS for token + snapshot writing
docs/guides/custom-agents.md	Adds/updates custom agents guide (including Python handler signature fix)
.github/workflows/cross-repo-issue-analysis.md	Adds agentic workflow definition (source markdown)
.github/workflows/cross-repo-issue-analysis.lock.yml	Adds compiled workflow lock file for CI execution
.github/aw/actions-lock.json	Adds gh-aw setup action pin for the new workflow

Comments suppressed due to low confidence (1)

go/internal/e2e/streaming_fidelity_test.go:151

• In the resume subtest, the original session is never destroyed before calling ResumeSession on a new client. Session.Destroy() is what releases resources and is the documented way to continue later via ResumeSession; not destroying here can leak resources and may prevent the session state from being properly finalized/persisted before resume. Please destroy session (and handle the returned error) before creating/resuming session2, and also ensure session2 is destroyed at the end of the subtest.

		session, err := client.CreateSession(t.Context(), &copilot.SessionConfig{
			OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
			Streaming:           false,
		})
		if err != nil {
			t.Fatalf("Failed to create session: %v", err)
		}

		_, err = session.SendAndWait(t.Context(), copilot.MessageOptions{Prompt: "What is 3 + 6?"})
		if err != nil {
			t.Fatalf("Failed to send message: %v", err)
		}

		// Resume using a new client
		newClient := ctx.NewClient()
		defer newClient.ForceStop()

		session2, err := newClient.ResumeSession(t.Context(), session.SessionID, &copilot.ResumeSessionConfig{
			OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
			Streaming:           true,
		})
		if err != nil {

[Copilot]

The sessions created in this test file are not destroyed at the end of the subtests. Other Go E2E tests call Session.Destroy() to release resources and avoid cross-test interference; leaving sessions alive until client.ForceStop() can cause resource leaks and make later subtests observe unexpected session/event state. Consider adding a defer/t.Cleanup that calls Destroy() (and checks/logs errors) for each created session.

This issue also appears on line 130 of the same file.

[Copilot]

The PR title/description describe a documentation-only fix in the custom agents guide, but this PR also adds/changes multiple E2E tests and snapshots across Node/Python/Go/.NET and introduces a new GitHub Actions workflow. Please update the PR title/description to reflect the broader scope, or split unrelated changes into separate PRs to keep review and rollback risk manageable.