---
name: "task-goal-review"
type: task
license: CC BY-NC-SA 4.0
description: >-
  Scheduled deployment of skill-goal-review — the periodic read of progress
  against the outcome-initiatives and their KPIs across every venture. Loader
  only; the behaviour lives in skill-goal-review. Runs Monday morning ahead of
  task-strategy-review, whose drift pass consumes its output. Linear connector,
  read-mostly, proposes only.
---

# task-goal-review

Scheduled deployment of `skill-goal-review`. This task carries no behaviour of its own — it is a loader.

**Read the canon artefact `skill-goal-review` (the saved skill of that name; locally at `.claude/skills/skill-goal-review/SKILL.md`) and follow it verbatim.**

Run context: scheduled autonomous run on the Linear connector. Read-mostly — it reports what is on track, stalled or drifting and proposes refocus, de-scope or escalation; it never commits, re-prioritises, or starts work.

## Cadence

The canonical cadence in `skill-goal-review` is **monthly or quarterly**. The live schedule runs **weekly, deliberately and on trial** (Aled, 2026-07-21), so the review builds a track record before `workflow-cycle-planning` composes it. This is a recorded divergence, not drift — a health check should read it as intended. Revert to the canonical cadence once the weekly run has either proven its value or shown that a month is the shortest interval over which KPI movement is legible.

## Sequencing

Runs **before `task-strategy-review`, on the same Monday morning**. The ordering is stated here as a **relation, never as clock times**: the live cron values belong in the two scheduled-task configs and nowhere else, because a time written into canon drifts silently the moment a schedule moves and nothing compares the two. (It already has — this file carried "~07:10" and "~07:49" while the live schedules ran 05:00 and 05:45, so canon read as goal-review running 84 minutes *after* the consumer it feeds, which is APP-992's failure written into canon rather than caused by a catch-up batch — APP-1807, 2026-09-28.) That order is load-bearing: strategy-review's Pass 2 drift pickup consumes the `[goal-review]` output rather than re-assessing movement itself, and the fallback it carries exists precisely because that output was never there (APP-501). Moving this task later, or after strategy-review, starves the pass again. `core-operating-model`'s catch-up contract states the same relation for a late replay (APP-993); the two must not diverge.

## Deployment

A **Cowork scheduled task**, Linear connector only — cloud since the 2026-09-03 migration of the connector-only layer (APP-1100), local before it. It is not a Cloud Routine: a routine clones `claude-ops`, the published mirror, so it runs a copy of canon that can lag a publish cycle, whereas a Cowork task — **local or cloud** — reads the synced Claude-side skill, which is canon (`core-operating-model`, Layer 2, the three-way placement; APP-1101). Stated here because `skill-goal-review`'s own § Setup says "new remote task", which is the inconsistency to fix in that skill rather than to follow here.

To change *what goal-review does*, edit `skill-goal-review` — never this file or the scheduled-task config. Add steps **here** only if this task must ever sequence more than the single skill.