Increase multi_agent_v2 default async wait timeout

[clkao]

Codex async idle monitoring now uses a five-minute per-call policy without changing Codex global configuration or launcher argv.

What changed

• State that wait_agent is asynchronous with respect to worker progress and captain interaction.
• Use wait_agent(timeout_ms: 300000) only when no captain-authorized active-scope work is ready.
• Require durable report/state verification followed immediately by the next-action loop.
• Retire obsolete Contractlint prose proxies; retain structural binding and filesystem checks.

Deliberately not included

• No synthetic delayed-worker timing harness or live timeout tests.
• No replacement phrase checks, aliases, configuration, or Claude/Pi/front-door changes.

Evidence

• Fresh independent validation passed go test -count=1 ./internal/contractlint (101 tests), full normal and race suites, formatting, and diff checks.
• The validated diff is the Codex adapter plus Contractlint removals only.

---

95w state record