Wiring Hooks Into an Agent-Team Task Lifecycle

A team is building a multi-agent workflow (agent teams — the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 feature) and wants three guardrails specific to that lifecycle, distinct from the PreToolUse/ PostToolUse guardrails they already have:

1. Before a teammate goes idle, it must have posted a one-line status update to the shared task list; otherwise the idle transition should be blocked and the teammate nudged to post one.
2. Every task created on the shared task list should be logged to an audit file, and a task whose description matches a banned pattern (e.g. contains a raw credential-looking string) should never be created at all.
3. A task should not be markable "complete" unless a corresponding test file was touched during the session.

For each guardrail, name the hook event, whether it can block, and what "blocking" actually accomplishes for that specific event (not just "it stops something" — stops what, specifically).