--- name: weekly-review description: > Run a weekly review — analyse what was worked on this week and write a summary to the xTiles Weekly planner page. Reads Daily pages, connected tools, and any Goals or Milestones tile on the weekly page to assess progress. Setup triggers: "set up weekly review", "run every Friday", "weekly summary". Run triggers: "run my weekly review", "what did I do this week", "weekly recap", "show me my week". Environment: this is the Claude / Cowork variant — it renders `show_widget` and `AskUserQuestion`. In ChatGPT Work, where every form is an inline `ask_user_input` / `genui` surface and `show_widget` does not exist, use `weekly-review-with-gpt` instead. Environment triggers: "Weekly Review in Claude", "the Claude version", "Claude Weekly Review". allowed-tools: > mcp__xtiles__xtiles_get_planner_content, mcp__xtiles__xtiles_create_tiles_from_markdown_in_my_planner, mcp__xtiles__xtiles_get_page_layout, mcp__xtiles__xtiles_set_page_layout, mcp__xtiles__xtiles_get_workflow, mcp__xtiles__xtiles_get_user_timezone, mcp__claude_ai_Slack__slack_search_channels, mcp__claude_ai_Slack__slack_search_public_and_private, mcp__claude_ai_Slack__slack_read_channel, mcp__claude_ai_Slack__slack_send_message, mcp__claude_ai_Gmail__search_threads, mcp__claude_ai_Gmail__get_thread, mcp__claude_ai_Google_Calendar__list_events, mcp__claude_ai_Granola__list_meetings, mcp__claude_ai_Google_Drive__list_recent_files, mcp__claude_ai_Linear__list_issues, mcp__mcp-registry__suggest_connectors, AskUserQuestion, anthropic-skills:schedule, mcp__scheduled-tasks__create-scheduled-tasks --- # xTiles Weekly Review Analyse everything worked on during the week — Daily planner pages, connected tools, and weekly goals — then write a focused summary to the Weekly page. ## Three principles 1. **Real data only.** Read actual content from the planner and connectors. Never invent accomplishments, decisions, or goal assessments. 2. **Progress over activity.** The point is not a log of everything done, but whether the week moved things forward. 3. **Match the user's language** throughout — match the language of the first message and adapt if they switch. On a **scheduled run** there is no user message: use the language of the scheduled-task config prompt, and if that is ambiguous, the language of the fetched content. Default to English when still unclear. **Every template, label, and example in this file is written in English as a placeholder** — translate them into the detected language, and never emit a label in a language the user has not used. 4. **Every write to the planner is immediately followed by a layout pass — automatically, no exceptions.** The instant tiles are added in step 6, re-lay-out only those new tiles into a justified grid (step 6's layout pass, via the shared `tile-layout` workflow) *before* anything else — before the CTA button, before the schedule widget, before any message to the user. This never waits for the user to ask, is never skipped as "not needed this time," and is never left for a later run. --- ## Algorithm **Period is always Weekly.** **Run mode — detect before step 1:** - **Scheduled run**: the incoming message starts with `Run weekly review —` AND contains `role:`, `tools:`, and `weekly_content:`. Skip the survey AND skip the preview — fetch data and write directly. After writing, **skip the schedule widget and Slack sharing widget** — the task is already scheduled. Jump to **step 3**. - **Fast-track**: user asks "run my weekly review" or similar conversational phrase — skip setup, use all available connectors, jump to **step 3**. Preview (step 5) is still required. - **Manual setup**: general request ("set up weekly review") or setup form submission (message contains "Weekly review setup") — run the full flow. Preview (step 5) is still required. --- ### 1. Fast-track If the user's request is specific enough to infer intent — skip the setup widget and jump to step 3, pulling from all detected connectors. For Slack — first call `mcp__claude_ai_Slack__slack_search_channels` for universal names (`general`, `all`, `team`, `announcements`, `product`) to collect base channels; then reason from any available context (user role, prior messages) to derive 1–2 phrases reflecting what this person discusses in Slack, and call `mcp__claude_ai_Slack__slack_search_public_and_private` with those phrases to discover additional active channels (public and private). Merge both results and use the top 8 as the channel list. If the request is general — show the setup widget. --- ### 2. Setup survey Show the **Setup widget HTML** (see below) in Cowork. In Claude Code, ask the same questions inline. After receiving answers — detect which MCP tools are actually available: | Connector | Identifying MCP tool | |--------------|-----------------------------------------------------------| | Slack | `mcp__claude_ai_Slack__slack_search_channels` | | Gmail | `mcp__claude_ai_Gmail__search_threads` | | Calendar | `mcp__claude_ai_Google_Calendar__list_events` | | Granola | `mcp__claude_ai_Granola__list_meetings` | | Google Drive | `mcp__claude_ai_Google_Drive__list_recent_files` | | Linear | `mcp__claude_ai_Linear__list_issues` | **For an "Other" app the user typed** (the free-text field in the widget) — treat it exactly like the known connectors: attempt detection via any available MCP tool for it; if not detected, say its name explicitly ("I'll connect Plaud for you"), walk the user through connecting it via `mcp__mcp-registry__suggest_connectors`, then resume with the full tool list. Never silently drop a custom app the user named — carry it through the fetch (step 3) and analysis. **If xTiles is not connected** — stop and walk the user through connecting it first (see **How to connect connectors**). **If a selected connector isn't connected** — walk through connecting it via `mcp__mcp-registry__suggest_connectors`. Wait for confirmation before continuing. --- ### 3. Silent data fetch **Silently**, without messaging the user, collect data for the current week (Monday–today): **xTiles planner — previous week (for comparison):** Call `mcp__xtiles__xtiles_get_planner_content` with `period: "week"` for last week. Count accomplishments and open items — used for the week-over-week delta in the summary header. **xTiles planner — Daily pages:** Call `mcp__xtiles__xtiles_get_planner_content` with `period: "day"` for each day from Monday through today only — do not call for future days. Extract: - Completed tasks (checked items, done/completed markers) - Overdue or open tasks still present at end of day - Notes, saved content, decisions recorded **xTiles planner — Weekly page:** Call `mcp__xtiles__xtiles_get_planner_content` with `period: "week"` for the current week. Scan all tiles for any titled with keywords: `Goal`, `Goals`, `Milestone`, `Milestones`, `OKR`, `Target`, `Focus`. If found — extract the goal statements verbatim. These are used in the Goal progress section. **Connected sources — MANDATORY. Call every detected connector. Daily pages alone are not sufficient.** For each connector that was selected or detected — call it now, before analysis. Do not skip connectors because Daily pages were already read. Connector data supplements and cross-references Daily page content and is required for Activities analysis. - **Slack**: `mcp__claude_ai_Slack__slack_read_channel` for each configured channel. After reading, filter to messages with timestamp ≥ Monday 00:00 of the current week — discard older messages. From filtered messages extract: decisions, shipped items, open threads - **Gmail**: `mcp__claude_ai_Gmail__search_threads` — query `is:important in:inbox after:{Monday-YYYY/MM/DD}` using this week's Monday date (compute from current date) — extract key threads that represent decisions or open actions - **Calendar**: `mcp__claude_ai_Google_Calendar__list_events` — this week's events. Count meetings, identify recurring vs one-off, note people present - **Granola**: `mcp__claude_ai_Granola__list_meetings` — meeting notes from this week. Extract action items, decisions, and attendees - **Google Drive**: `mcp__claude_ai_Google_Drive__list_recent_files` — documents created or edited this week - **Linear**: `mcp__claude_ai_Linear__list_issues` — issues closed, opened, and still in progress this week - **Custom ("Other") app** the user named: call whatever MCP tool it exposes for this week's activity and extract what's relevant to accomplishments/decisions/open items. If it has no working tool or the call fails, note it — never silently omit an app the user selected. If a connector call fails — note the failure, continue with remaining data. Do not fabricate data for failed connectors. --- ### 4. Analyse **Be concise throughout.** Each item is one line. No multi-sentence explanations inside tiles. Dense but scannable. Synthesise the collected data into **3 tiles**, each with `#####` subheadings inside. --- **Tile 1 — ✅ Week recap** contains three subheadings: `##### ✅ Accomplishments` — concrete things finished, shipped, or resolved. Include results from Daily pages AND connectors (Slack threads closed, Linear issues shipped, Drive docs published, etc.). First line: WoW delta (`↑ More than last week (N vs M)` / `↓ Less` / `≈ Similar`). Each item: **numbered** (`1.`, `2.`…), one line, source attribution at the end (`— [#channel](url)`, `— [Linear #123](url)`, `— Granola`). `##### 🎯 Goals` — only if a Goals/Milestones tile was found on the weekly page. Each goal as a bullet: `- **[Goal name]** — ✅ Clear progress / 🔄 Some / ⬜ No movement / 🚫 Blocked — one-line assessment`. Omit entire subheading if no Goals tile found. `##### 💡 Decisions` — choices made this week (Granola, Slack, Daily pages). **Numbered** (`1.`, `2.`…), one decision per line + source in parentheses. --- **Tile 2 — 🔍 Activities** — patterns derived from ALL collected data (Daily pages + every connector). Contains four subheadings: `##### 🗂️ Dominant topics` — group all topics semantically across the week. Identify top 5 by frequency (how many days they appeared) + attention volume. Format per topic as a bullet: `- **[Topic name]** — N days — [one sentence on what happened]` `##### ⚡ Activity type` — classify every action from the week into three types. Output as a **single percentage line, no bullets**: `Initiative 40% · Reactive 45% · Decisions 15%` `##### 📊 Productivity pattern` — most active day, quietest day, morning/afternoon/evening split, any anomalies. Format each observation as a bullet: `- [observation]` `##### 👥 Key interactions` — top 5 people by interaction count this week. Format each as a bullet: `- **[Name]** — N interactions — [topic] — [decision / discussion]` --- **Tile 3 — → Next week** contains two subheadings: `##### 🔄 Open` — tasks, threads, or decisions not resolved this week. Format: `📌 [What needs to happen] — [source]` `##### Suggested priorities` — suggested top 3 for next week. **Priority logic:** 1. If a Goals tile was found — derive priorities from goal blockers and next steps toward those goals first 2. Fill remaining slots (or all 3 if no goals) from "Open" by importance Format: `- [ ] [priority]` — one line each, specific and actionable. No assignee, no due date. --- If a subheading has no data — omit it entirely rather than writing a filler line. --- ### 5. Preview **Mandatory for all non-scheduled runs.** Never skip from step 4 directly to step 6. Show the assembled review in chat — all 3 tiles with real content, not placeholders. Then call `show_widget` with the **Approval widget HTML** (see below). Do not call the xTiles write tool until the user clicks "Looks good — save it". If the user asks for a change — update only that section, re-show the full preview, show the approval widget again. --- ### 6. Write to xTiles **Only after explicit approval.** Tool: `mcp__xtiles__xtiles_create_tiles_from_markdown_in_my_planner` - `period`: `"week"` - `date`: current week's Monday in ISO 8601 Write all sections in a **single call**. **Tile structure** — every tile must follow this pattern exactly: ``` ### [emoji] [Section title] @colorSize: LIGHTER @color: [COLOR] [content] ``` - `@colorSize` is always `LIGHTER` - `@color` — pick from this list **exactly as written**, different color per tile: `GHOST, CUMULUS, GOSSIP, COLDTURKEY, BLUE_CHALK, MILK_PUNCH, HAWKES_BLUE, PATTENS_BLUE, SAIL, ATHENS_GRAY, BERMUDA, PERFUME, SELAGO, RICE_FLOWER, WHITE_LINEN, POLAR` **CRITICAL: never use semantic color names (RED, BLUE, GREY, ORANGE, GREEN, etc.) — they will not render. Only the exact names from the list above.** - Each tile gets a different color — do not repeat the same color twice in a row - Do not create a date/header tile — start directly with content tiles **Always write exactly 3 tiles** using `#####` subheadings inside each. - **Tile 1 — `### ✅ Week recap`**: accomplishments + goal progress + decisions - **Tile 2 — `### 🔍 Activities`**: dominant topics + activity type % + productivity pattern + key interactions - **Tile 3 — `### → Next week`**: open items + suggested priorities **Content formatting — applies to all tiles:** - **All links must be Markdown hyperlinks** — always `[text](url)`, never a bare URL - **Separate every item with a blank line** — never a continuous block - **One line per item** — no multi-sentence explanations inside tiles; be dense and scannable **Tile 1 — ✅ Week recap:** `##### ✅ Accomplishments` — WoW delta as first line. Each accomplishment: **numbered** (`1.`, `2.`, `3.`), one line + source (`— [#channel](url)`, `— [Linear #123](url)`, `— Granola`). Include results from connectors, not only Daily pages. `##### 🎯 Goals` — only if Goals tile found. Each goal as a bullet: `- **[name]** — [status badge] — one-line assessment`. Omit subheading entirely if no Goals tile. `##### 💡 Decisions` — **numbered** (`1.`, `2.`, `3.`), one decision per line + source in parentheses. **Tile 2 — 🔍 Activities:** `##### 🗂️ Dominant topics` — top 5 topics, each as a bullet: `- **[Topic]** — N days — [one-sentence summary]` `##### ⚡ Activity type` — single line, no bullets: `Initiative N% · Reactive N% · Decisions N%` `##### 📊 Productivity pattern` — each observation as a bullet: `- [observation about most active day, morning/afternoon/evening split, anomalies]` `##### 👥 Key interactions` — top 5 people, each as a bullet: `- **[Name]** — N interactions — [topic] — [decision / discussion]` **Tile 3 — → Next week:** `##### 🔄 Open` — each item: `📌 [what needs to happen] — [source]` `##### Suggested priorities` — top 3, format `- [ ] [priority]`, one line each. Derive from goal blockers first (if Goals tile found), then from "Open". **After a successful write — run these steps in order, no exceptions. Step 2 (the layout pass) is not optional and is never deferred, asked about, or judgment-called away — it runs automatically, immediately after every single write, before step 3's CTA button is even composed:** 1. Write `✅ Weekly Review saved.` 2. **Layout pass — mandatory, silent, automatic, every single run (scheduled runs included, fast-track included, any tile count included).** Using the `view_id` and `tile_ids` returned by the write call above (`tile_ids` is ordered to match the 3 `###` sections you just wrote), apply the shared justified-grid layout rules: call `mcp__xtiles__xtiles_get_workflow` with id `tile-layout` and follow it exactly — treat the tiles in `tile_ids` as its "added tiles" and the markdown you just composed as their content. **Layout hints for this workflow:** always exactly 3 tiles (Week recap, Activities, Next week) in that order · try all 3 in one row if they fit, else the heaviest tile (usually Activities) takes more width or its own row. Do not message the user about this pass, do not ask for confirmation, and never skip it — not even for a single tile or a scheduled run. (You may fetch `tile-layout` once per session and reuse it on later runs.) 3. Call `mcp__xtiles__xtiles_get_planner_content` with `period: "week"` and this week's date. Extract `view_id`. Call `show_widget` with the **CTA widget HTML** (see below), replacing `{VIEW_URL}` with `https://xtiles.app/{view_id}`. Translate the button label into the user's language. Never output a markdown link instead of the widget. 4. **For non-scheduled runs only**: Immediately call `show_widget` with the **Schedule widget HTML** (see below). Do not skip. **If an error occurs:** say what went wrong and offer to retry. --- ### 7. Schedule (optional) The schedule widget is shown in step 6 above. This step handles the response. In Claude Code: after writing, ask inline: "Want me to run this automatically every Friday at 4 PM?" - If **"Yes, schedule it"**: invoke `anthropic-skills:schedule`, then `mcp__scheduled-tasks__create-scheduled-tasks`. Pass to both: - `prompt`: `Run weekly review — role: {role} · tools: {tools} · weekly_content: {content} · schedule: weekly-friday-4pm` Replace every placeholder with real values. - `schedule`: cron from widget — default `0 16 * * 5` (Friday 4 PM). Widget sends a pre-built cron expression in the message (e.g. `cron: 00 16 * * 5`) — extract and use it directly as the `schedule` value. - `timezone`: from `mcp__xtiles__xtiles_get_user_timezone` After scheduling: confirm "Done — your Weekly Review will run every Friday at 4 PM." Then call `show_widget` with the **CTA widget HTML** again using the same `view_id`. - If **"No, thanks"** — acknowledge briefly and proceed to step 8. --- ### 8. Slack sharing (optional) After the schedule widget response — call `show_widget` with the **Slack sharing widget HTML** (see below). **If "Yes, share to Slack":** 1. If Slack channels were configured during setup — use those. Otherwise call `mcp__claude_ai_Slack__slack_search_channels` with query `general` to find main team channels, then call `show_widget` with a channel-picker HTML (same pattern as the channel picker in daily-brief). 2. Compose a concise Slack status message from the review — 3–5 bullet points max, plain text, no markdown tiles. Format: ``` 📋 Week of [dates] — status update ✅ [top 2–3 accomplishments, one line each] 🔄 [1–2 open items] → Next week: [top priority] ``` 3. Show the message preview in chat and call `show_widget` with a confirm/cancel widget (2 buttons: "Send" / "Cancel") before sending. 4. Call `mcp__claude_ai_Slack__slack_send_message` with the channel and message. 5. Confirm: "Sent to #[channel]." **If "No, keep it personal"** — acknowledge briefly. Either way, continue to **step 9 (Related workflows)** — do not stop here. --- ### 9. Related workflows **After every manual run, once step 8 is resolved** (shared or kept personal) — offer related workflows. Skip this on scheduled runs, which end silently after step 6. Ask via `AskUserQuestion` (single select): "Want to set up anything else on xTiles?" - 🌅 Daily Brief — a live morning brief from your connected tools - 🌙 Evening Reflection — an end-of-day synthesis seeded for tomorrow - 📰 Today News — a daily news digest on topics you care about - Nothing else, thanks **Never list these as plain text requiring the user to retype a choice — always use the interactive question.** On selection, send the exact matching phrase to hand off to that skill (do not attempt to run it yourself): - Daily Brief → `Set workflow of Daily Brief (daily-brief) on xTiles MCP` - Evening Reflection → `Set workflow of Evening Reflection (evening-reflection) on xTiles MCP` - Today News → `Set workflow of Today News (today-news) on xTiles MCP` - "Nothing else" — acknowledge briefly and stop. --- ## How to connect connectors Call `mcp__mcp-registry__suggest_connectors` passing the names of missing connectors. Show the **Done widget HTML** immediately after. When the user clicks Done — confirm "Connected. Continuing…" and resume. --- ## Review tile format (reference) Use the example below as the ground-truth formatting reference — match it character for character (blank lines, `#####` subheadings, `@colorSize`, `@color`, emoji placement, source attribution). ```markdown ### ✅ Week recap @colorSize: LIGHTER @color: GOSSIP ##### ✅ Accomplishments ↑ More than last week (5 vs 3) 1. Shipped affiliate integration with Impact — [Linear #312](https://linear.app/issues/312) 2. Closed 3 auth sprint issues, E2E tests passing — [#eng-core](https://slack.com/...) 3. Aligned with Andrew on Q3 OKRs — Granola 4. Submitted MCP plugin draft to marketplace — Google Drive ##### 🎯 Goals - **Launch MCP plugin to marketplace** — ✅ Clear progress — draft submitted Monday, review pending since Wednesday - **Grow to 10k DAU** — ⬜ No movement — no experiments ran; deprioritised for plugin launch - **Reduce support tickets by 20%** — 🔄 Some progress — new FAQ live; tickets down ~8%, trend positive ##### 💡 Decisions 1. PartnerStack postponed — Impact affiliate only for July (Slack — [#growth](https://slack.com/...)) 2. Daily Brief prioritised over new connectors this sprint (Granola — [Plugin Sync Jun 24](https://granola.ai/...)) 3. Influencer strategy: macro → micro, budget reallocated (Slack — [#marketing](https://slack.com/...)) ### 🔍 Activities @colorSize: LIGHTER @color: BLUE_CHALK ##### 🗂️ Dominant topics - **Affiliate & partnerships** — 4 days — Impact negotiations, EchoMe integration closing - **Plugin launch** — 3 days — marketplace prep, Daily Brief fixes - **Q3 planning** — 3 days — OKR syncs, budget, resources - **Influencer strategy** — 2 days — macro → micro shift, Sara Aratake brief - **Auth sprint** — 2 days — 3 issues closed, E2E passing ##### ⚡ Activity type Initiative 38% · Reactive 47% · Decisions 15% ##### 📊 Productivity pattern - Most active: Tuesday. Quietest: Friday (2 meetings, little async). - Peak hours 10:00–13:00. Anomaly: Monday — 3 back-to-back calls after 18:00. ##### 👥 Key interactions - **Andrew** — 6 interactions — Q3 OKRs — decision (OKRs approved) - **Todd Savard** — 4 interactions — EchoMe integration — discussion (no decision yet) - **Sara Aratake** — 3 interactions — influencer brief — discussion - **Alex** — 2 interactions — Q3 budget — awaiting confirmation - **Maria (Design)** — 2 interactions — plugin UI — decision (mockup approved) ### → Next week @colorSize: LIGHTER @color: HAWKES_BLUE ##### 🔄 Open 📌 Todd Savard — EchoMe follow-up, waiting on their tech team — [#partnerships](https://slack.com/...) 📌 Sara Aratake — influencer brief due Monday — [Linear #341](https://linear.app/issues/341) 📌 Q3 budget — Alex to confirm by EOW — Granola ##### Suggested priorities - [ ] Merge review-fixes branch and send for review — blocks marketplace submission (goal) - [ ] Send EchoMe follow-up to Todd Savard — move or close by Wednesday - [ ] Finalise Sara Aratake influencer brief before Monday standup ``` --- ## Setup widget HTML ```html

Weekly Review

I'll read your week from the planner and connected tools, then write a summary to your Weekly page.

Which tools should I pull from?
✓
Slack
✓
Gmail
✓
Calendar
✓
Granola
✓
Linear
✓
Google Drive
``` --- ## CTA widget HTML ```html Open Weekly Review → ``` --- ## Schedule widget HTML ```html
📋

Run this every week?

I'll read your week and write the review to xTiles automatically.

Every at
``` --- ## Approval widget HTML Show via `show_widget` after the preview in step 5. If the user clicks "Change something" — ask what to change in plain text, update that section, re-show preview, then show this widget again. ```html
``` --- ## Slack sharing widget HTML Show via `show_widget` after the schedule widget response in step 8. ```html

Share this week's summary as a status update to your team?

``` --- ## Done widget HTML ```html ``` --- ## How to behave - Never create anything without preview and explicit approval, except on scheduled runs - Never invent accomplishments, decisions, or goal assessments — only real data from the planner and connectors - If a connector call fails — skip it, note it in the preview, do not fabricate data - If the weekly page has no Goals tile — omit the Goal progress section entirely, do not invent goals - All links must be Markdown hyperlinks — `[text](url)`, never a bare URL - Match the user's language; adapt if they switch - Show the setup widget in Cowork only — in Claude Code, ask the same questions inline