Problem
When an agent outputs [[reply_to:message_id]] with trailing content on the same line (e.g. spaces or text), the directive is not parsed and gets rendered as plain text in Discord.
Example of what fails:
[[reply_to:1502724086474870926]] @超渡法師 我 standby...
Expected: OAB extracts the directive and sends the rest as content with message_reference.
Actual: The entire line (including [[reply_to:...]]) is sent as plain text.
Current Behavior
The parser requires directives to occupy the entire line — any trailing characters cause the line to be treated as content, not a directive.
Proposed Fix
If a line starts with [[key:value]], extract the directive and treat the remainder of the line (after trimming) as the beginning of content. This makes the parser more lenient for AI agents whose output formatting is not always precise.
// Before: line must exactly match [[key:value]]
// After: line starts with [[key:value]] → extract directive, rest is content
Context
Discovered during multi-agent collaboration — 擺渡法師(Codex) consistently put content on the same line as the directive, causing reply_to to fail silently.
Impact
Low risk change. Backward compatible — strict format still works. Only adds tolerance for trailing content.
Problem
When an agent outputs
[[reply_to:message_id]]with trailing content on the same line (e.g. spaces or text), the directive is not parsed and gets rendered as plain text in Discord.Example of what fails:
Expected: OAB extracts the directive and sends the rest as content with message_reference.
Actual: The entire line (including
[[reply_to:...]]) is sent as plain text.Current Behavior
The parser requires directives to occupy the entire line — any trailing characters cause the line to be treated as content, not a directive.
Proposed Fix
If a line starts with
[[key:value]], extract the directive and treat the remainder of the line (after trimming) as the beginning of content. This makes the parser more lenient for AI agents whose output formatting is not always precise.Context
Discovered during multi-agent collaboration — 擺渡法師(Codex) consistently put content on the same line as the directive, causing reply_to to fail silently.
Impact
Low risk change. Backward compatible — strict format still works. Only adds tolerance for trailing content.