Fix pr-create and pr-status command reliability

[tombakerjr]

Bug 1: pr-create fails when no PR exists yet

Location: commands/pr-create/SKILL.md (or equivalent)

Problem: The skill template runs gh pr view to check for an existing PR before creating one. This command exits non-zero when no PR exists (the expected/common case). The Bash pattern wrapper treats this as a fatal error and stops execution instead of interpreting it as "no PR exists, proceed to create."

Fix: Either:

• Use gh pr view 2>/dev/null || true and check output
• Use gh pr list --head $(git branch --show-current) --json number --jq '.[0].number' which returns empty instead of erroring

---

Bug 2: pr-status doesn't wait for CI or review comments

Location: commands/pr-status/SKILL.md (or equivalent)

Problem: Current implementation is a point-in-time snapshot. It doesn't:

1. Watch CI checks to completion (gh pr checks --watch)
2. Wait for the Claude review comment to be posted (which arrives ~10-12s after CI passes)
3. Verify the review comment is from the most recent CI run (by timestamp comparison)

Expected behavior:

1. Run gh pr checks --watch to block until CI completes
2. After CI passes, wait briefly then fetch issue comments
3. Find the Claude review comment (from claude[bot]) — it always posts something (suggestions or LGTM)
4. Verify the comment timestamp is after the latest CI run started (ensures it's not a stale comment from a previous push)
5. Report any blockers: 🔴 CRITICAL, 🟡 FIX, BLOCKER
6. Never indicate "ready to merge" without finding the review comment from the current run

Context: Discovered during ATLAS PR #97 audit remediation workflow. Both bugs required manual workarounds.