@windwhiterain

Per-model outbound proxy routing for DeepSeek Harness: send only the models you name through a proxy, with a direct fallback while it is unavailable.

A persistent office of colleague agents for DeepSeek Harness: ordinary sessions with a roster, channels, direct messages, and delivery that wakes a colleague when something is addressed to it.

Hide the model-facing skill loader for the agents of the preset that mounts it, so a session can load a skill only when the user invokes it explicitly with /name.

Agent-facing recall of a session's own pre-compaction history for DeepSeek Harness: an outline, a paginated transcript read, and a literal search over the durable log.

Named subagent templates for DeepSeek Harness: a template fixes a starting route (or the route pool it resolves one from), an agent preset, and an optional persona, so a delegating agent picks by name.

Per-session automatic compaction threshold for DeepSeek Harness: a compaction backend whose trigger percentage is stored per session and adjustable from the Web composer while the session runs.

Answers an exhausted DeepSeek Harness account quota: move the session to another route of its pool when one still has allowance, and otherwise retry the same request every interval with no attempt limit.