Revert tool-choice forcing — the Codex regression was strict:null

The GPT-account (Codex) path supports tool calls and worked before; my
forced tool_choice broke it and was the wrong fix. The real regression was
strict:null in the rebuild's tool definitions, which made the endpoint drop
the tool list so the model had nothing to call — a "Done" run that hired no
one. Forcing then turned that silent failure into a red one.

Restore the original "auto" loop (no forced tool_choice, no finish tool) and
keep only the one-line strict:null omission. Codex now receives valid tools
on the working "auto" choice, as in the version that worked.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

AgentCompany-New/src/main/services/agent-tools.ts

@@ -119,16 +119,8 @@ export const AGENT_TOOLS: ToolDefinition[] = [
       required: ["title", "doc_type", "content"],
     },
   },
-  {
-    name: "finish",
-    description: "End your turn. Call this ONLY after you have taken every action this task needs — for a kickoff that means hiring your team, delegating their first tasks, and posting a kickoff message to a channel. Do not call it just to stop early.",
-    parameters: { type: "object", additionalProperties: false, properties: {}, required: [] },
-  },
 ];
 
-/** Control tool that ends an agent's turn; the run loop consumes it, never dispatched. */
-export const FINISH_TOOL = "finish";
-
 export function dispatchAgentTool(deps: AgentToolDeps, ctx: AgentToolContext, call: ToolCall): string {
   try {
     switch (call.name) {