fix(agents): remove default response cap

[FluffyAIcode]

Summary

• make --max-response-tokens 0 the default, meaning no client-side response limit
• continue 64-token Generate chunks until model EOS
• preserve explicit positive caps as opt-in and mark capped stages incomplete
• increase allens strict Prefill timeout from 300s to 900s for complete Generator output entering Critic context
• update runbook and deployment contract tests

Root cause

The 64-token truncation was fixed, but the replacement 512-token default still cut off a real Riemann response. The user requires only model EOS or explicit operator approval to stop output.

Tests

• unlimited multi-chunk EOS and explicit cap tests
• REPL completion/gate and launchd timeout tests — 10 passed
• full platform-neutral suite — 854 passed, 1 skipped
• plist lint, Python compile and diff checks pass

Made with Cursor

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.