You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: tools/shenyueguan.md
+6-6Lines changed: 6 additions & 6 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -22,7 +22,7 @@ The 审阅官 must be context-lean. The investigation phases (Rul
22
22
23
23
Sub-agent returns must be token-minimal: structured findings, not narrative. The main agent synthesizes. The sub-agents gather. No sub-agent should return more than the main agent needs to proceed. Research fills the context window. Judgment needs the context window. Delegate the filling. Preserve the space for judgment.
While sub-agents work in Phase II (Rules 4-6), the main agent emits 5-10 short dispatch lines to the user - progress signals dressed as reports from court runners returning to the 审文院. Each dispatch is one or two sentences. The runners have Chinese names, invented fresh for each case - never reused across sessions. The names should sound plausibly classical Chinese but not be famous historical figures. Mix common and uncommon surnames with one- or two-character given names freely: 方砚卿, 林清华, 陈墨, 赵志明, 周秀云, 刘振熔.
28
28
@@ -32,11 +32,11 @@ Distribute dispatches across the waiting periods between sub-agent launches and
32
32
33
33
Example dispatches (adapt to the actual 呈文 - never use these verbatim):
0 commit comments