[chuks-qua]

What

Mcode thread tools used to show their raw names in the timeline, such as thread_target_list. A call with no arguments also showed {} beside the name. A decided tool approval read "Allowed once mcp__mcode__thread_target_list".

Now each thread tool row has a plain label, such as "Listed targets". This holds for running, finished, failed, and grouped rows. A row whose input is empty shows no detail. A tool approval receipt names the tool the same way its row does, both live and after reload.

I verified it with a real supervised Claude turn in Electron that called thread_target_list. The captures below show the same thread on the old and new renderer. Targeted web tests pass (105 files, 1099 tests), and typecheck and oxlint are clean. The new tests failed on the old code (54 failures).

Closes #2270

Why

People who direct agents read these rows to follow what the agent did. Raw MCP names and a stray {} make that harder to read, and the receipt disagreed with the row above it.

UI Changes

Before. The row shows the raw name and {}, the group reads "1 thread_target_list call", and the receipt shows the prefixed name.

After. The row reads "Listed targets", the group reads "Listed targets 1 time", and the receipt reads "Allowed once Listed targets".

The running row also reads "Listed targets" while it waits for approval.

Review Notes

• The pending approval card still shows the exact tool name and input. That card is where the user decides, so the precise name matters there.
• thread_create_batch gets no label. Only the legacy external transport exposes it.
• A tool with no usable input detail no longer falls back to showing its own name as the detail. One existing test pinned that fallback, and it now expects the label alone.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}