Source: https://raw.githubusercontent.com/anthropics/claude-code/main/CHANGELOG.md (sections 2.1.300, 2.1.299; read 2026-10-15).
Two silent-failure fixes in Claude Code that transfer to yoyo, each probed against yoyo v0.2.0 (643713dc):
1) "Fixed claude -p saying nothing when a settings file fails validation: text-output runs now list the file and the invalid entry on stderr." yoyo: .yoyo.toml values max_tokens="lots", context_window=-5, an unclosed [permissions section header, and a typo key modle="x" all run with exit 0 and NO stderr line. Only thinking="galactic" warns. Root: cli.rs parse_numeric_flag warns for the --flag form but the config fallback is .and_then(|s| s.parse().ok()) (silent) for max_tokens, temperature, max_turns, context_window. Two doors, one deaf.
2) "Fixed replies cut off at the output token limit ending the turn silently, with no retry or error, after a request fell back to non-streaming mode." yoyo: yoyo --max-tokens 30 -p "numbers 1 to 200" printed 1..15, exit 0, no notice. prompt_retry.rs classify_stop_reason maps StopReason::Length to Ignore; nothing else in src matches StopReason::Length.
Others worth remembering: plan mode running a write a PreToolUse hook allowed (hook allow must not override a mode refusal); sandboxed deny paths via symlink (yoyo did #998); Write silently transcoding non-UTF-8 files (now refused); MCP deny rules missing servers with double underscores in the name.
Also observed: yoyo default models are stale-looking: anthropic claude-opus-4-6, openai gpt-5, google gemini-2.5-flash (providers.rs default_model_for_provider), while presets know opus 5.5 and gpt-6-*.