Update protocol, docs-discovery skill, and doc layering
- Switched AI AGENT CORE PROTOCOL to new `[LOADED_DOCS: N files (+K this turn: ...)]` prefix format in all primary instruction files; updated examples and enforcement details.
- Added `docs-discovery` skill for proactive .md doc loading before code search; documented usage and integration as a shared agent skill.
- Introduced `## Protocol History` section and `LLM_PROTOCOL_DECISIONS.md` decision log for cross-repo protocol change tracking.
- Expanded protocol-audit skill and `REPOS.md` to support 8 instruction files (primary/inherit split), new invariants, and known issues.
- Added/updated structured TODO and issues docs for serialization, logging, and SignalR binary protocol.
- Improved cross-references, doc layering, and clarified documentation-first coding workflow.
- Various minor doc clarifications and formatting for protocol consistency.
Refactored AcBinaryHubProtocol and AyCodeBinaryHubProtocol to remove the VarUInt length prefix for raw byte[] arguments. Now, the protocol writes a tag (0x44) followed directly by the raw bytes, with argLength implying the payload size. Updated read logic to match: on detecting the tag, the code skips it and returns the remaining bytes as the payload. Updated documentation to clarify the new fast-path, protocol roles, and AcBinary detection. Set BufferWriterChunkSize to 4096 for SignalR in the base protocol for better alignment with Kestrel. Marked the related issue as resolved.
- Rewrote SequenceBinaryInput for lazy TryGet iteration (no segment array allocation), zero-copy access to segment backing arrays, and efficient cross-boundary reads using a reusable ArrayPool scratch buffer.
- Added Release() to IBinaryInputBase; now always called after deserialization to return scratch buffer.
- BufferWriterChunkSize is now mutable; set to 4096 for SignalR protocol for better pipe alignment.
- Added and updated documentation: detailed input buffer lifecycle, cross-boundary handling, and new BINARY_ISSUES.md and SIGNALR_ISSUES.md for known limitations and planned optimizations.
- No breaking API changes; improves performance, memory usage, and diagnostics for multi-segment binary deserialization.
Loretta