docs(tend): simplify statusline cache check

[max-sixty]

Summary

#3900 now excludes diagnostic collector commands from profiles, so the weekly skill no longer needs instructions for discounting those commands. This removes that workaround and the historical 29/32 subprocess threshold, which was a one-time observation that had already drifted and could not distinguish useful new work from a regression.

The weekly cache check now focuses on duplicate (command, context) operations. The profile still exposes command_count for targeted performance investigation, and the recipe comment now describes only the duplicate data that jq .cache returns.

Checks

• pre-commit run --files .claude/skills/running-tend/SKILL.md
• git diff --check

This was written by Codex on behalf of max-sixty

[worktrunk-bot]

Checked the deletion's own stated precondition and it has landed: the collector's subprocesses carry a reserved context (DIAGNOSTIC_CONTEXT in src/trace/emit.rs), and profile_entries filters it out ahead of both command_count and CacheReport::from_entries. Running the recipe here, the render's git worktree list --porcelain [.] and the collector's [(diagnostic)] copy land in separate buckets, and command_count reports 16 against 18 traced commands — so the Measurement artifact bullet described behavior the code no longer has, and the discount clause would have double-subtracted three.

One consequence worth naming: the ~29/~32 baseline was the section's only signal for newly added distinct subprocesses. A change that adds five fresh git calls to the render produces no duplicate (command, context) pair, so the remaining check reports nothing. Since command_count is already collector-free, keeping the baseline would have cost only the deletion of the "minus three" clause — so dropping it is a scope call rather than a forced one.