daily-fix: Step 10d re-snapshots Guard-1 strip on merge conf

Overview / Motivation

Auto-filed by the /daily 2026-07-08 problem sweep (transcript-mined; emitting agent: /daily orchestrator).

Goal

Make the Step 10d merge robust to fleet marker churn between the Guard-1 snapshot and the merge.

Workflow gap

• Bug observed: Transcript 78a4234b (issue-1128) ~08:48-09:22Z: 'GraphQL: Pull Request has merge conflicts (mergePullRequest)' then 'MERGE FAILED (non-lint transport failure)'; completion note 09:24Z records two worktree recovery merges with foreign tasks/ ultimately pinned to a single captured main SHA.
• Why it is a workflow gap: The recovery recipe (pin to one captured main SHA) is only discovered as the SECOND fallback; under heavy concurrent marker commits the first-attempt snapshot is routinely stale.
• Confidence (emitter): medium

Proposed change (candidate diff sketch — refine in planning)

Step 10d merge-guard block: capture MAIN_SHA once, strip foreign tasks/ against that SHA, and on a mergePullRequest-conflict error re-snapshot against a freshly captured SHA and retry once (documented, not silent).

Scope / surfaces

• Primary target: .claude/skills/issue/SKILL.md
• Grep the workflow surface for the pattern before editing and update every hit; list them in the plan.

Constraints / invariants

• Workflow-surface only — never experiment code, configs/, or tasks/.
• scripts/workflow_lint.py --check-asks + --check-references pass; ruff on touched files passes.
• This session runs under EPM_WORKFLOW_FIX_SESSION=1 and carries a workflow_fix_target: Provenance line — it MUST NOT auto-route its own subagents' workflow-fix candidates (recursion guard).

Provenance

• workflow_fix_target: .claude/skills/issue/SKILL.md
• origin: transcript-mined by /daily 2026-07-08 problem sweep
• evidence: mine-D P3 (78a4234b ~08:48-09:24Z)