You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

you have been detected be spamming with unwarranted prs and issues and your issues and prs have been automatically closed. please read contributing guide Contributing.md.

Greptile Summary

Replaces the overly permissive prefix + suffix regex pair in isSilentReplyText() with a single whole-string match (^\s*TOKEN\s*$), fixing a bug where substantive replies ending with NO_REPLY were silently suppressed (issue #19537). The old suffix regex \bNO_REPLY\b\W*$ would match the token at the end of any response, even after real content. The old prefix regex ^\s*NO_REPLY(?=$|\W) similarly matched tokens followed by additional content. Both are now correctly replaced with a strict whole-string check.

• Bug fix: Only responses consisting entirely of the silent token (with optional surrounding whitespace) are now treated as silent, preventing legitimate replies from being dropped.
• Tests added: New colocated test file src/auto-reply/tokens.test.ts with 7 cases covering exact match, whitespace variants, and the regression scenario.
• Dead code removed: isSilentReplyPrefixText was also removed — it had no callers in the codebase.

Confidence Score: 5/5

• This PR is safe to merge — it's a targeted bug fix with correct logic and good test coverage.
• The change is small and well-scoped: a single regex replacement in one function with clear before/after behavior. The new regex logic is correct — it only matches whole-string tokens. Tests cover the key edge cases including the specific regression. The function's callers all use it in a way compatible with the tighter matching. No removed code had any callers.
• No files require special attention.

_{Last reviewed commit: 8ed852d}

Added follow-up fixes beyond the original regex tightening:

• Integrated streaming partial-token suppression work from fix: suppress silent-reply partial tokens during streaming #19648 into this PR (guarded isSilentReplyPrefixText path).
• Applied prefix suppression in streaming/typing flow paths (including heartbeat token partials in runner normalization).
• Updated regression tests so only exact NO_REPLY is suppressed; substantive text like NO_REPLY -- ... now delivers.
• Added prefix-helper coverage and NO_/NO_RE streaming partial regression coverage.
• Rebased onto latest main and resolved merge conflicts in agent-runner-execution.ts and typing.ts.

Current head: 203ec4f6cc2a4c58a6621439d808c0151438dc6d.

Landed via temp rebase onto main.

• Gate: pnpm test src/telegram
• Land commit: d75e2ec
• Merge commit: 97fa44d

Thanks @aldoeliacim!