The Cadence layer at week resolution: close the loop on the week, surface what stalled, and aim the next week at the priorities that matter.
Not to be confused with: /audit (is the kit healthy — connections, freshness) or /level-up (build one new automation). This skill reviews your week and priorities.
When to run
- End of week (retrospective) or start of week (planning) — it does both.
The output (always this shape)
## Weekly review — week of <date>
**Moved** — progress on each 90-day priority this week (concrete, not vibes).
**Stalled / at risk** — priorities or loops that didn't move + why + the unblock.
**Closed loops** — commitments resolved this week.
**Carrying forward** — open loops into next week, with owners.
**Next week's focus** — 3–5 outcomes for the coming week, tied to priorities.
**Surfaced for /level-up** — any task done manually 3+ times this week (candidate to automate).
Process
- Read:
context/priorities.md,decisions/log.md(this week's entries),projects/, and open loops (tools/open_loops.pyif present). - Assess each priority: moved / stalled / at-risk, with evidence (a shipped thing, a decision, a metric) — not assertion.
- Diagnose stalls: for each, name the single blocker and the unblocking action/owner.
- Plan next week: propose 3–5 outcomes mapped to the 90-day priorities; keep it realistic.
- Spot recurring manual work → list it under "Surfaced for /level-up".
- Offer to log: append a dated summary entry to
decisions/log.md(allowed without approval).
Autonomy
L2 — drafts the review and next-week plan; the human confirms.
Guardrails (from CLAUDE.md)
- Evidence over assertion — "verify before you claim done."
- No confidential figures persisted; reference live sources.
- Route next-week owners to the team where appropriate.