[bokelley]

Summary

Adds a short Parallel worktrees subsection to .agents/playbook.md, under "Local Hooks". Git worktrees isolate the working tree, index and HEAD, but they share the object store, every ref (including origin/*) and the stash. So one session's git fetch moves origin/main under every other worktree of the same clone.

The rule it adds:

• Resolve the base SHA once, after the first fetch, and use that SHA for rebase, diff and test baselines. Never re-resolve origin/main mid-task.
• Force-push with an exact lease, --force-with-lease=<branch>:<expected-sha>. A bare lease compares against the shared origin/<branch>.
• Never use bare git stash / git stash pop. Prefer a WIP commit, or a tagged stash restored by SHA.

Why

On 2026-11-08 a session rebasing #8512 re-resolved origin/main partway through its task. Another session's fetch had just moved it to #8798, so the commit landed on top of #8798 and would have reverted three of its files. The session caught it before pushing. Several agents now work in parallel worktrees of one clone, so this will recur without a written rule.

Notes

• Agent-tooling docs only. No protocol surface, so no changeset; check-changeset-protocol-scope passes.
• .agents/** is a sensitive path, so it needs a human merge.

🤖 Generated with Claude Code

[aao-secretariat]

Ladon verdict: Escalate to human review

Escalate — gated path requires human/CODEOWNERS approval.

This PR modifies .agents/playbook.md, which matches the gated-path glob .agents/** (gated_paths: true), and review_decision is REVIEW_REQUIRED — not APPROVED. Per decision rule row 2, this is a hard, non-overridable approval gate: the outcome cannot be approve or plain comment until a real human/CODEOWNERS approval lands.

The diff itself is clean: a single docs-only change adding a "Parallel worktrees" section (20 additions, one file). The reviewer verified the technical claims and found no Critical/High/Medium findings. No protocol surface, schema, wire, or changeset implications. Absent the gated-path gate this would approve — but the gate is unsatisfied.

Escalation reasons

• .agents/playbook.md is under the .agents/** hard approval gate; requires human/CODEOWNERS review (review_decision is REVIEW_REQUIRED, not APPROVED).

Why human review

• Modifies .agents/playbook.md which matches the gated path .agents/**; human/CODEOWNERS approval required (review_decision is REVIEW_REQUIRED, not APPROVED).
• This PR touches a path under a hard, non-overridable approval gate (.agents/playbook.md (modified) matches .agents/**) and the current GitHub review decision is 'REVIEW_REQUIRED', not APPROVED. This is a hard gate enforced in code — Ladon cannot auto-approve until a human/CODEOWNERS approval is recorded, regardless of how clean the diff is.