Release 0.13.4

[github-actions]

Release readiness review (v0.13.3 -> TARGET f3afa6a)

This is a release readiness report done by $final-release-review skill.

Diff

v0.13.3...f3afa6a

Release call:

🟢 GREEN LIGHT TO SHIP Patch release contains targeted fixes, version bump is aligned, and no concrete release-blocking issues were found.

Scope summary:

• 8 files changed (+206/-37); key areas touched: src/agents/extensions/models/, tests/models/, .github/scripts/ + PR label schema/prompt, and version metadata (pyproject.toml, uv.lock).

Risk assessment (ordered by impact):

1. AnyLLM replay input sanitization broadens field stripping

• Risk: 🟡 MODERATE. Sanitization now recursively removes None values and adapter-only fields, which could alter edge-case replay payload semantics for providers expecting explicit nulls.
• Evidence: src/agents/extensions/models/any_llm_model.py replaces _remove_openai_responses_api_incompatible_fields with _sanitize_any_llm_responses_input and recursively strips provider_data, fake IDs, and None; tests updated in tests/models/test_any_llm_model.py.
• Files: src/agents/extensions/models/any_llm_model.py, tests/models/test_any_llm_model.py
• Action: Run targeted replay-path validation and keep green if passing: uv run pytest -s tests/models/test_any_llm_model.py -k "sanitize or replay"; pass criteria: all selected tests pass and no provider regression is reported in CI.

2. PR label taxonomy migration (feature:lite-llm -> feature:extensions)

• Risk: 🟢 LOW. Internal automation behavior changed but is covered by corresponding script, schema, prompt, and tests in the same diff.
• Evidence: Label constants and inference rules changed in .github/scripts/pr_labels.py, schema/prompt updated, and tests revised/added in tests/test_pr_labels.py.
• Files: .github/scripts/pr_labels.py, .github/codex/schemas/pr-labels.json, .github/codex/prompts/pr-labels.md, tests/test_pr_labels.py
• Action: Run label tests and keep green if passing: uv run pytest -s tests/test_pr_labels.py; pass criteria: test suite passes with expected feature:extensions inference.

No material risks identified.

Notes:

• Base tag was selected from local tags only: v0.13.3.
• Target was current HEAD: f3afa6aa749dae90226cbe85116c574026a7890d.
• Commit span reviewed: 5e0e6d26, 40aada1a, f3afa6aa.
• Assumed standard CI verification for target commit; no additional full local verification stack was executed in this review.