fix(workflow): stop e2e-slot mangling the stack session name

[iscekic]

Problem

stack_session() sanitises a worktree basename with tr -c 'A-Za-z0-9_-' '_'. basename emits a trailing newline, the newline is outside the preserved set, so tr rewrites it to _ — and $(...) can no longer strip it. Every session name came back with a spurious trailing underscore:

$ basename /Users/igor/Projects/.worktrees/mobile-ui-ddc7 | tr -c 'A-Za-z0-9_-' '_'
mobile-ui-ddc7_          # wanted: mobile-ui-ddc7

Why it matters

The slot/stack unification promises that release frees the slot and stops the worktree's stack, so a stack can never outlive the slot that entitled it. That guarantee was not holding. Three silent failures:

1. release never stopped the dev stack. stop_stack looked up kilo-dev-<name>_, tmux has-session failed, and it returned early treating the stack as already gone. Stacks outlived their slots — the exact over-subscription the cap exists to prevent.
2. status always reported stack=none, even with the stack plainly up.
3. stack_is_covered never matched on the recorded worktree, so coverage worked only via the owner-name-prefix fallback. A slot whose owner name does not begin with the worktree basename would make a live stack look uncovered — and stacks --reap would then pnpm dev:stop a running E2E.

Observed on a live host: slot-1 … [/Users/igor/Projects/.worktrees/mobile-ui-ddc7 stack=none] while tmux has-session -t kilo-dev-mobile-ui-ddc7 exited 0.

Fix

Keep \n in the preserved set so tr leaves the newline alone for $(...) to strip. Other unsafe characters are still replaced (weird.name/x → weird_name_x).

Verification

Against live slot state, before and after:

before: slot-1: … [/Users/igor/Projects/.worktrees/mobile-ui-ddc7 stack=none]
after:  slot-1: … [/Users/igor/Projects/.worktrees/mobile-ui-ddc7 stack=up]

bash -n clean. stacks still reports every running stack as covered, now on the worktree path rather than by accident of the fallback.

Found while enforcing the new slot rules across running dev stacks after #4826.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

The one-line fix correctly keeps \n in the tr preserved set so the trailing newline from basename is stripped by $(...) instead of being rewritten to _, restoring correct behavior for stop_stack, status, and stack_is_covered.

Files Reviewed (1 files)

• .kilo_workflow/e2e-slot.sh

---

_{Reviewed by claude-sonnet-5 · Input: 12 · Output: 2.4K · Cached: 200.8K}

_{Review guidance: REVIEW.md from base branch main}