Reserved-Token Rejection Recheck

Verdict

RESERVED_TOKEN_REJECTION_NEEDS_FIX

The new rejection of TAB, LF, CR, NUL, backslash, and reserved sentinels is accepted.

Remaining gap

The proposed document_id grammar permits aliases such as dot segments, empty segments, and repeated separators. Distinct textual IDs can therefore identify the same logical document, violating SSOT and Điều 14.

Required correction

• Reject . and .. segments.
• Reject empty segments and //.
• Require exact equality with the canonical document ID returned by MCP.
• Apply the same canonical identity rule to both sides of every set comparison.
• Require marker literals to match the exact extractor grammar and require marker_kind consistency; a broad ^<!--.*-->$ rule is insufficient.