Release readiness review (v0.6.6 -> TARGET 112b242)

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

Diff

v0.6.6...112b242

Release call:

• GREEN LIGHT TO SHIP Incremental changes with tests and docs updates; risks are low and scoped.

Scope summary:

• 136 files changed (+8200/-2275); key areas touched: experimental Codex extension, handoff history/run observability, shell tool output truncation, compaction session, docs (incl. translations), examples.

Risk assessment (ordered by impact):

1. Handoff history filtering now omits summarized tool/function call items from forwarded input.
   • Risk: MODERATE. Downstream agents relying on raw tool/function items in handoff input could see reduced context compared to previous behavior.
   • Files: src/agents/handoffs/history.py, src/agents/_run_impl.py, src/agents/run.py
2. Compaction deferral when local tool outputs exist.
   • Risk: LOW. Compaction may be delayed, which could increase session size until a forced compaction occurs.
   • Files: src/agents/run.py, src/agents/memory/openai_responses_compaction_session.py
3. Shell tool now truncates output based on max_output_length.
   • Risk: LOW. When callers set limits (or executors return them), outputs are shorter than before; could affect consumers that expect full stdout/stderr.
   • Files: src/agents/_run_impl.py
4. New experimental Codex extension and tool.
   • Risk: LOW. Additive and under experimental namespace, but introduces a large new surface area; ensure experimental status is clear to users.
   • Files: src/agents/extensions/experimental/codex/codex_tool.py, src/agents/extensions/experimental/codex/*.py, examples/tools/codex.py

Notes:

• BASE_TAG v0.6.6 and TARGET 112b242d31ff80d59cda1fe4e8430f96b39571b4 derived from local tags and HEAD.
• No material risks identified beyond the low/moderate items above; no local verification was run here.