[folathecoder]

Summary

Fixes the nine issues identified during the Aldwin 0.0.1 pre-release audit and end-to-end PTY review so ENG-385 (Release Aldwin 0.0.1) can begin cleanly. Turns that finish via Agentling's built-in final_answer tool now render their full Markdown answer without showing an internal final_answer tool card, approval prompts wait indefinitely without being cancelled at 60 seconds by Agent.tool_timeout, long tool rows unwrap JSON before truncation and keep duration · summary visible at 80 columns, and \+Enter and Ctrl+J insert newlines.

Ticket

Closes ENG-405

• Dev plan: ENG-406
• Test plan: ENG-407

Changes

• Host (src/aldwin/host/runner.py):
  • Suppress final_answer from ToolCallEvent and non-error ToolResultEvent in AldwinHost._run_turn, and emit TextEvent from final_answer when the final step did not stream TextDelta chunks (FR-1).
  • Unwrap primary string fields (command, url, name, query, text) in format_args_preview before truncating at 120 characters (FR-3).
• Core tools (src/aldwin/core/tools/run_command.py, src/aldwin/core/tools/persona_tool.py):
  • Set timeout: float | None = float("inf") on RunCommandTool and SetPersonaTool so Agent.tool_timeout (60.0 s) does not cancel ask() while waiting for user approval, leaving RunCommandTool's post-approval 60-second timer in control (FR-2).
• CLI (src/aldwin/cli/):
  • Cap display_args width in TurnRenderer.on_tool_result so duration · summary is never clipped on 80-column terminals, and add a truncated-JSON fallback in format_display_args (FR-3, FR-4).
  • Bind Ctrl+J (c-j) and \+Enter (c-m) to insert newlines in build_prompt_session and update /help keybindings (FR-5).
  • Validate OPENAI_API_KEY from os.environ on first run via needs_onboarding / run_onboarding (FR-6).
  • Add PLAIN_PROMPT_STYLE (noreverse on bottom-toolbar) for --plain and NO_COLOR=1 (FR-7).
  • Remove (FR-1) from setup and uninstall --help docstrings, cap display_home in the session header Panel title so panels stay <= 76 columns, italicize continue in colour-mode notices, and remove unused short_session_id (FR-8).
• Tests (tests/):
  • Added 13 unit and integration regression tests (TC-01 to TC-15) across test_host_runner.py, integration/test_full_turns.py, test_run_command.py, test_persona.py, test_cli_render.py, cli/test_render_suite.py, test_cli_chat.py, test_cli_onboarding.py, and test_cli_main.py.

How it was tested

• Automated:
  • uv run ruff check src tests (passed)
  • uv run ruff format --check src tests (69 files formatted)
  • uv run mypy src tests (0 issues in 65 source files)
  • uv run pytest --cov=aldwin --cov-report=term-missing --cov-fail-under=90 (246 passed, 2 deselected in 4.19s, 92.49% coverage)
• Manual / PTY (TC-16):
  • Built wheel with uv build, installed into an isolated UV_TOOL_DIR, and verified aldwin --help, 80-column long tool rows, final_answer Markdown rendering, \+Enter and Ctrl+J multi-line input, and --plain mode (0 reverse-video SGR codes).

Notes for reviewers

• Agent.tool_timeout remains 60.0 for non-approval tools (ENG-380 FR-1), while RunCommandTool and SetPersonaTool opt out of the pre-approval wrapper via timeout = float("inf").