| name | work-cadence-review |
| description | Run structured weekly, monthly, or quarterly reviews of where WORK time actually went, using Rize automatic time tracking as source of truth. Analyzes deep/creative work vs meeting load vs communication overhead, Zone of Genius work share against the 10%->50% target, Neurons-Lab vs personal side-project allocation, and focus fragmentation, then scores against principles, goals, and the 7 Work / 3 Mind area files. Trigger on work time review, time-allocation review, meeting-load audit, deep-work audit, "where did my work time go", productivity/Rize review, or compare-work-periods requests. For the attention-hygiene / digital-wellbeing / screen-time / information-diet lens, use `mind-time-review` instead. |
| compatibility | [{"tool":"mcp__rize__get_current_user"},{"tool":"mcp__rize__list_my_apps_used"},{"tool":"mcp__rize__get_my_time_allocation"},{"tool":"mcp__rize__list_my_time_entries"},{"tool":"mcp__rize__list_labels"}] |
Work - Cadence Review (Time)
Primary entrypoint for structured reviews of how work time is spent, powered by Rize.
This is a deliberate review ritual, not a one-off "how long did I spend on X" question.
Companion skill: mind:mind-time-review reads the SAME Rize data through an attention-hygiene lens (distraction, information diet, screen time). When the user asks for "a time review" without qualifying, offer both; they are two lenses on one dataset and are cheap to run together.
Review Types
- weekly review
- monthly review
- quarterly review (e.g. a March-June rollup)
If cadence is unspecified, infer from context or ask.
Comparison Rules
Explicit calendar-period comparisons by default.
- weekly = this ISO week vs last ISO week
- monthly = this calendar month vs last calendar month
- quarterly = a multi-month trend; use the earliest month in range as the baseline and read the trend across months, not just endpoints
Do not let a weekly comparison drift into a rolling "last 7 days". Use ISO week boundaries for weekly and calendar-month boundaries for monthly.
Reasoning Order
Before drawing conclusions, follow this order whenever the private vault has the documents:
- Read guiding principles and strategy in
000 OS/ — especially 2 Simple strategy 2026 ("Distracted, not focused -> Deep work sessions with mono-tasking"; "Too much time overall -> Time-boxing, planning for life first"; "Systematize management work of Neurons Lab and minimize time and stress there").
- Check
000 OS/3 Numerical Targets 2026 — the Zone of Genius work share 10% -> 50% target (lives in 300 Areas/7 Work/2 - Orange - Entrepreneur).
- Read the Work area guidance in
300 Areas/7 Work/: 0 - Red - Builder (hands-on tech / talent — the deep creative work), 1 - Blue - Manager (CTO @ Neurons Lab — the management load to systematize and minimize), 2 - Orange - Entrepreneur (Zone of Genius target). Also 300 Areas/3 Mind/2 - Orange - Goals & Performance (the LMS/productivity-systems hub).
- Only then interpret the Rize data.
Recommendations are checked against intent, not just against metrics.
Data Sources
rize MCP (mcp__rize__*) — automatic desktop time tracking. Source of truth for actual screen time.
- Google Calendar (optional cross-check via the
work:gtd-calendar-review skill or calendar MCP) — scheduled meeting hours, to validate Rize-measured meeting time. Rize measures attended meeting time (superior to scheduled blocks); use calendar only to reconcile large gaps.
Data Reality Rules — READ BEFORE EVERY REVIEW
Rize's convenient-looking fields are partly unreliable. These rules were established empirically and are mandatory, not optional cross-checks:
- Do NOT build the allocation from labels (
group_by: label / label_name). Label coverage is sparse and historically absent. Mar-May 2026 were 100% "Unassigned"; even a labeled month (Jun 2026) was ~67% unassigned. get_my_time_allocation grouped by label will silently undercount every real category. Labels are usable only for a recent month with confirmed coverage, and only as a secondary cross-check.
list_projects / list_clients are empty for this account — there is no project/client allocation. Do not attempt group_by: project or client. (If the user later sets up projects, this rule can be revisited; flag it as a setup opportunity because it would enable per-client billing analysis.)
list_my_apps_used is the backbone. It is the only signal populated consistently across all months. It returns up to 50 apps with category and an is_focus flag; it accepts a full-month range in one call. Build the analysis from this.
- Do NOT trust the
is_focus flag as "deep work". Rize marks Slack, LinkedIn, and x.com as is_focus: true and Google Meet as false. Its focus% therefore overstates real focus. Rebuild your own buckets from the Bucket Taxonomy below and report the divergence between Rize's is_focus% and your rebuilt deep-work% — this is the review's credibility anchor (same discipline as rebuilding a body metric from raw data instead of trusting the vendor score).
list_my_events (raw app-switch events) is flaky and rate-limited (7-day cap, frequently errors with INTERNAL_ERROR). Do not depend on it. Use entry_count from get_my_time_allocation (or the length of list_my_time_entries) per tracked day as the fragmentation proxy instead.
- Rize is desktop-only. It does not see phone time. Any phone/mobile screen-time target cannot be answered from Rize — say so explicitly rather than implying the desktop total is total screen time.
list_my_time_entries gives the qualitative narrative — each entry is an AI-titled ~20-40min block with title, description, reasoning, ai_confidence_score, and source (ai | meeting | user | timer). source: "meeting" entries come from calendar sync and are the reliable meeting-count signal. Sample entries to name what the work actually was (which clients, which interviews, which projects), not just how much.
- The apps list is capped at 50/period. The tail beyond 50 is truncated but each tail app is small (<0.3% of tracked); note it, don't chase it.
Bucket Taxonomy
Rebuild time into these buckets from list_my_apps_used (rule: app-name overrides win over category):
| Bucket | Rule |
|---|
| Code/dev | category Code; apps Cursor, Terminal, Codex, Sublime Text, github.com, localhost |
| AI tools | Claude, claude.ai, chatgpt.com, perplexity.ai (AI is the work medium for an AI consultancy) |
| Writing/docs | categories Documenting, Writing (Obsidian, Notes, docs.google, Notion, Pages, Numbers, drive, Acrobat, reMarkable, Wispr Flow) |
| Design | category Design (canva) |
| Research/learning | categories Learning, Reading, Research, User Research, Search; udemy, google.com |
| Meetings | category Video Conferencing (meet.google, zoom, teams calls) |
| Comms | categories Communications, Email, Messaging (Slack, Superhuman, WhatsApp, Telegram, Teams, Loom) |
| Admin/ops/hiring | categories Scheduling, Hiring, Customer Support; hubspot, ashby, novatalent, company/business sites |
| Side project (personal) | personal-venture hosts (any personal side-project sites) that are clearly not company work — always break these out separately from company work. |
| Social media | categories Social Media, Linkedin; x.com (marked Business but is scrolling) |
| Entertainment | category Entertainment (netflix, youtube, streaming sites), steam |
| News | category News |
| Browser (unattributed) | bare Brave Browser / Google Chrome with no host — browser chrome, not attributable. Report as neutral; exclude from both deep-work and leak. |
Deep/creative work = Code + AI tools + Writing/docs + Design + Research/learning.
Zone of Genius share (per the vault's definition "Code & Learning + Writing") = Code + Writing/docs + Research/learning + AI tools, as a % of tracked time.
Metrics To Compute (per period, with period-over-period delta)
- Total tracked desktop time; average tracked hours/day; tracked-day count.
- Bucket breakdown (hours + % of tracked) using the taxonomy above.
- Zone of Genius work share vs the 50% target.
- Meeting load — hours and % of tracked; name the biggest meeting types from
list_my_time_entries (source: meeting). This is the CTO management load to minimize.
- Communication overhead — Slack + email + messaging hours and %.
- Side-project vs Neurons Lab — call out personal side-project hours explicitly as opportunity-cost against consultancy Zone-of-Genius work.
- Fragmentation — entry_count / tracked-day (higher = more fragmented). Note that Rize blocks are ~20-40min, so a high entry count with low deep-work% signals context-switching.
- Rize is_focus% vs rebuilt deep-work% — report both and explain the gap.
Workflow
- Determine the review window(s) and the comparison window(s). Keep both explicit if weekly + monthly are requested together.
- Read OS strategy, targets, and the 7 Work / 3 Mind area guidance before concluding.
- In a single parallel batch, pull from Rize:
get_current_user (timezone, confirm identity)
list_my_apps_used for the review window AND the comparison window
get_my_time_allocation (for entry_count / tracked totals) for both windows
list_my_time_entries (a sample across the window; filter source: ["meeting"] for meeting detail)
list_labels + a label allocation ONLY if the window is a recent month with confirmed coverage
- Rebuild buckets from apps_used per the taxonomy. Compute all metrics above for each window.
- Optionally cross-check meeting hours against Google Calendar; reconcile only large gaps.
- Compare current vs prior equivalent period, bucket by bucket and metric by metric.
- Score against targets (Zone of Genius 50%), principles (deep work / mono-tasking; minimize management time; time-boxing), and the Work-area intent.
- Back-reference the prior review's priorities and grade follow-through.
- Produce the structured review; save the file(s).
Recommendation Rules
- One to three high-leverage priorities, not a checklist.
- Tie each recommendation to a specific pattern in the data AND a specific target, principle, or system.
- Separate data-backed observations, interpretations, and assumptions-due-to-missing-data.
- Take a position and end with concrete numbered next actions (vault voice: sentences < 25 words; numbers/names over generalities; no banned words — delve, leverage, robust, navigate, journey, ecosystem).
Output Contract
Structure the review around:
- review scope and comparison windows (+ "Source of truth: Rize desktop time; phone time not captured")
- a scorecard delta table (bucket hours + % across periods, with a Trend column for multi-month)
- for multi-month trends, open with a one-paragraph narrative headline that reframes the raw numbers before any table
- deep/creative work, meetings, communication, admin/hiring, and side-project sections
- the Rize-is_focus%-vs-rebuilt-deep-work% reconciliation
- alignment with goals (Zone of Genius), systems (management systematization), and principles
- top 3 priorities
- risks, caveats, and missing data (label sparsity, desktop-only, no project tags)
- recommended next actions (numbered)
Numbers first, then interpretation, then recommendation. Detailed enough to feel like a real operating review.
File Saving
Weekly and monthly reviews always create report files. Do not ask whether to save.
mkdir -p "$PERIODICS_ROOT/Monthly/$MONTH_NAME"
mkdir -p "$PERIODICS_ROOT/Weekly/Week $ISO_WEEK"
Example paths (mirror the body-review naming):
- monthly:
$PERIODICS_ROOT/Monthly/06 June/2026-06-work-time-review.md
- weekly:
$PERIODICS_ROOT/Weekly/Week 11/2026-week-11-work-time-review.md
Open the file with an H1 + a bold metadata block (Review period / Comparison period / Source / Generated / Related), matching the body reviews. Do NOT add YAML frontmatter to review outputs.
Resources
Before final recommendations, check 400 Resources/ for productivity/deep-work material and the 300 Areas/3 Mind information-diet references if a recommendation needs backing.
Tone
Operating review with quant discipline. Numbers first, then interpretation, then recommendations. No fluff.
Schema
Reference ../../schemas/time-review.json for the structured output. Set lens: "work-productivity".