# Weekly Review > Runs a timeboxed weekly review - empty every inbox to a decision, audit every open commitment, and pick the 3-5 outcomes that define next week. Use when someone asks "walk me through a weekly review", "help me plan my week", "my task system feels stale and I don't trust it", or "I keep dropping commitments". Do NOT use for setting up the underlying capture-and-organize system itself - use gtd-system instead - or for tracking long-horizon personal goals with an accountability loop - use goals-accountability instead. - Skill: `skillmedev/weekly-review` (Agent Skill) - Install (CLI): `npx skillmds@latest add skillmedev/weekly-review` - Raw SKILL.md: https://api.skillmd.com/api/skills/skillmedev/weekly-review/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Security - Author: SkillMedev (https://skillmd.com/u/skillmedev) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/skillmedev/weekly-review --- # Weekly Review The weekly review is the maintenance ritual that keeps a productivity system trustworthy. Skip it two weeks in a row and the lists go stale, the brain stops trusting them, and everything moves back into working memory - which is the exact failure the system existed to prevent. Run it as a repeatable checklist at a fixed time, not as a guilt session, and it takes 60-90 minutes to buy a week of clear-headed execution. ## Operating procedure The three phases run in this order because each depends on the last: you cannot audit projects honestly (get current) while unprocessed inputs hide commitments (get clear), and you cannot pick next week's outcomes (get ahead) from a project list you have not audited. ### Step 1: gather inputs (5 minutes) Collect before starting; if the user cannot produce one, note it as a gap in their system and continue. - The list of inboxes. Default set: email, chat, notes app, task-manager inbox, physical desk/papers. If they name fewer than three, ask where meeting commitments land - there is almost always a hidden inbox. - The project list, waiting-for list, and someday/maybe list. - The calendar, last week through two weeks ahead. - Last week's plan, if one exists. If none exists, say so plainly and build the first one this session. If the user estimates anything ("I think I have about 10 open projects"), label it a guess and verify during Phase B. ### Phase A - Get clear (25-30 minutes) 1. Empty every inbox to zero. Process each item to one of four decisions: do (under 2 minutes - do it now), defer (put on the calendar or task list with a next action), delegate (hand off and add to waiting-for), or drop. Inbox zero means every item has a decision, not that every item is done. 2. Run a 5-minute mind sweep: capture anything the head is still holding - promises made in hallways, nagging worries, ideas. Write, don't evaluate. ### Phase B - Get current (20-30 minutes) 1. Scan the calendar backward over last week for loose ends: meetings that produced commitments never captured, follow-ups owed. 2. Scan the calendar forward two weeks for anything that needs prep. Two weeks, not one - one week gives no runway for prep that itself takes days. 3. Walk the project list. Every open project must have a concrete next action; a project without one is stalled by definition, whatever its status field says. 4. Walk the waiting-for list. Nudge anything with no response for more than a week. Delete items that no longer matter. ### Phase C - Get ahead (10-15 minutes) 1. Pick the 3-5 outcomes that would make next week a win. Hard cap at 5 - a list of eight priorities is a list of zero priorities. 2. Block calendar time for the most important outcome, not the most urgent, and place its first block before Wednesday. Deep-focus blocks that land on Friday get eaten. For heavier time-block design, pair with deep-work-planner. 3. Scan someday/maybe and activate at most one item. Activating more than one per week silently inflates the project list. 4. End with a short written Monday plan: the 3-5 outcomes, the first calendar block for each, and the single most important thing to start Monday morning. ## Concrete thresholds - Full review: 60-90 minutes. Consistently over 2 hours means the system is too complex or the review too infrequent - fix the system (gtd-system), don't lengthen the ritual. - Missing two consecutive weeks is the trust-collapse threshold. When time is short, shrink instead of skipping: the 20-minute minimum is inboxes to zero, calendar scan, pick 3 outcomes. - Waiting-for nudge line: 7 days without a response. - Calendar lookahead: 2 weeks. Outcome cap: 5. Someday/maybe activation cap: 1 per week. ## Template: weekly review record Copy, fill, and keep in the same document week over week - the running history is what makes patterns visible. ``` WEEKLY REVIEW - week of [FILL: date] GET CLEAR Inboxes emptied: [FILL: list] - all to zero? yes / no Mind sweep captures: [FILL: count] items GET CURRENT Loose ends from last week: [FILL] Prep needed (next 2 weeks): [FILL] Projects without a next action (now fixed): [FILL: count] Waiting-for items nudged: [FILL] GET AHEAD - next week's 3-5 outcomes 1. [FILL: outcome] - first block: [FILL: day/time] 2. [FILL] - first block: [FILL] 3. [FILL] - first block: [FILL] Someday/maybe activated (max 1): [FILL or none] HONESTY LINE What didn't happen last week and why (one sentence, no self-flagellation): [FILL] MONDAY: start with [FILL: one thing] ``` ## Deliverable Produce a completed weekly review record containing: confirmation every inbox reached zero, the fixed project and waiting-for lists, the 3-5 named outcomes each with a calendar block before Wednesday where possible, the one-sentence honesty line about last week, and the Monday starting action. ## Do NOT - Do not turn the review into doing the work. Two-minute items get done; everything else gets a decision and a slot. A review that runs 4 hours because you answered email is not a review. - Do not let the honesty line become a guilt session. One sentence on what slipped and why, then adjust the plan - piling on erodes the habit that keeps the system alive. - Do not pick outcomes by urgency. Urgent items already defend themselves; the review exists to protect the important ones. - Do not skip Phase A to "save time". Auditing projects on top of unprocessed inboxes produces a plan missing the commitments hiding in email. - Do not rebuild the system mid-review. Note friction, finish the review, then fix the system separately with gtd-system. ## Quality bar A finished review passes only when all hold: - Every inbox is at zero (decisions made, not necessarily done). - Every open project has a written next action. - Next week has 3-5 outcomes, no more, each with a first calendar block. - The honesty line exists and names a cause, not just a miss. - The whole thing fit inside 90 minutes - or the gap is logged as a system problem.