Make qa-explore default to dangerous mode for automation

ID	INFRA_TOOLING-MAKE_QA_EXPLORE_DEFAULT_TO_DANGEROUS_MODE_FOR_AUTOMATION
Status	completed
Priority	p1
Milestone	m1-ft-analytics-analyst-pilot
Owner	sr-engineer-architect
Completed by	dave
Completed	2026-03-18

Problem

Default qa-explore to dangerous mode, keep a safe opt-in, add tests, and update docs so worker QA does not stall on Playwright permissions.

scripts/dispatch already defaults worker automation to dangerous mode and offers --safe as the explicit opt-in for permission prompts.
scripts/qa-explore still defaulted to safe mode, which caused inner claude -p browser runs to stop and ask for MCP Playwright tool permissions instead of completing automatically.
The follow-up goal is to align qa-explore with the existing dispatch automation model, keep a documented safe escape hatch, and cover the behavior in tests/scripts/test_dispatch_scripts.py.

Flip scripts/qa-explore to dangerous-by-default, add an explicit --safe flag, and update docs/tests accordingly. Recommended because it matches dispatch, removes prompt brittleness in worker automation, and keeps the interactive path available when deliberately requested.
Leave the default alone and rely on prompts to tell workers to pass --dangerous. Rejected because it is brittle and makes the wrapper itself unreliable for automation.

Change scripts/qa-explore so its default permission mode is dangerous and --safe becomes the explicit override.
Update the dry-run and execution tests in tests/scripts/test_dispatch_scripts.py to lock the new behavior in.
Update the repo QA guidance to document the new default and the meaning of --safe.
Validate with focused script tests and record the result here.

Updated scripts/qa-explore usage/help text and argument parsing to default to dangerous mode, mirroring scripts/dispatch.
Replaced the old --dangerous opt-in with --safe as the only permission-mode override.
Updated tests/scripts/test_dispatch_scripts.py so dry-run now asserts Permission mode: dangerous by default, asserts --safe removes the dangerous flag, and asserts real execution includes --dangerously-skip-permissions.
Updated .codex/skills/qa-explorer/SKILL.md and AGENTS.md to document the automation-first default and the --safe opt-in.
Focused validation: uv run pytest tests/scripts/test_dispatch_scripts.py -q -> 15 passed.

N/A. This task changes the local QA wrapper and docs, not a user-facing browser surface. Focused script regression tests cover the behavior change.
[x] QA exploration completed (or N/A for non-UI tasks)