explain-worklisted

# explain-work — review coupled with "explain this to me" Take one unit of work and explain it at the altitude the available context justifies. The value is not new analysis — it is **composing** analysis you already produce (structural, ownership, intent) and **teaching** it. This is a generic skill: it contains no personal or org-specific data. All situated context is read at runtime from sources below, so the skill stays shareable and PII-free. **Output is local only. Post nothing, edit nothing.** ## Arguments `$ARGUMENTS` — `<PR number, owner/repo#N, LINEAR-ID, or path> [--deep]` - The target resolves to a PR, a Linear ticket, or a path/diff (auto-detected). - `--deep` runs the full review-lens panel (see Step 4). Default is a light pass. ## Step 0 — Check for prior session context Before fetching anything, check whether the current session already contains analysis of the same target. If a `pr-review-kit`, `pr-review`, or `review-prep` pass was just run in this session for the same PR/path, **reuse that context rather than re-fetching**. The layers that are already materialised (file reads, commit history, inline comment triage, matrix rows) should be treated as the `hard` layer from Step 2 — no need to repeat the gh calls or worktree setup. Signal that session context was reused in the degradation report (Step 6) so the user knows the explanation was built from the live analysis, not a cold fetch. ## Step 1 — Resolve the unit of work and authorship 1. Detect t