dcurlewis/.claude/skills/shutdown

Interactive end-of-day shutdown: consolidate today's outcomes, surface unresolved items, and prepare carryover context for tomorrow's /morning.

Pre-Requisites

• Today's daily journal should exist (Curated-Context/Daily-Journals/YYYY-MM-DD.md), typically created by /morning
• Meeting insights for today are optional but enriching

Process

Step 0: Context Gathering (silent)

Display "Gathering today's context..." then collect:

0a. Fetch and process today's Slack activity. Display "Fetching today's Slack activity..." then:

1. Run /slack-fetch (read and execute .claude/skills/slack-fetch/SKILL.md) in daily mode (no arguments).
2. If /slack-fetch wrote files to Raw-Materials/Slack/, run /slack (read and execute .claude/skills/slack/SKILL.md) to process them into a summary.
3. Note the resulting summary file path for use in Step 0f context reading.
4. If /slack-fetch fails or returns no messages, note it and continue. Do not block on this step.

0b. Read today's journal (Curated-Context/Daily-Journals/YYYY-MM-DD.md). If it does not exist, warn the user and ask whether to proceed with limited context or create a minimal journal first.

0c. Check for unprocessed meeting transcripts in Raw-Materials/Meeting-Transcripts/. Count and note filenames. These will be surfaced in Step 1.

0d. Check other raw material queues (Raw-Materials/Docs/). Count files per non-empty queue. These will be surfaced in Step 1.

0e. Consolidate today's action items. Scan Curated-Context/Meeting-Insights/ for files in the current month's subdirectory (YYYYMM/) that start with today's date (YYYYMMDD). For each matching file, extract ## My Action Items bullets. Build a consolidated list attributed by meeting title.

0f. Read today's journal sections (### Notes, ### Agenda, ### Weekly Tasks, ### Slack Highlights) and, if a new Slack summary was produced in Step 0a, read that summary file for additional context.

0g. Fetch tomorrow's calendar. Run the calendar connector with the tomorrow flag (default: python3 Scripts/connectors/calendar-today.py tomorrow --output Synced-Data/Calendar/tomorrow.json). See Guidelines/data-sources.md for alternative calendar sources. Read the resulting file. If the script fails or no calendar connector is configured, note that tomorrow's calendar is unavailable and skip tomorrow-related questions in Step 2.

0h. Read memory index (Curated-Context/System/memory-index.md) for current priorities.

Step 1: Surface Pre-Interview Flags

Before the interview, present anything requiring attention:

• Unprocessed transcripts: If any exist in Raw-Materials/Meeting-Transcripts/, list filenames and ask: "Process these via /meeting before reflecting? (y/n)". If yes, run /meeting (read and execute .claude/skills/meeting/SKILL.md) for each transcript, then re-scan for action items (repeat Step 0d) and add new items to the consolidated list. If no, proceed.
• Other queue items: Note counts from other queues (informational only, no blocking).

If nothing to flag, skip directly to Step 2.

Step 2: Interactive Interview

Present the consolidated action item summary first, then ask 3-5 contextual questions in a single message. Target: 3-5 minutes total from the user.

Action item summary (always show):

**Today's action items** (from N meeting insights):
- [Meeting Title]: [item]
- [Meeting Title]: [item]

If no meeting insights were processed today, show: "No meeting insights processed today."

Questions (contextually adapted, not a fixed checklist):

| # | Always? | Template | Adaptation rule | |---|---------|----------|----------------| | 1 | Yes | "What felt like the most important thing today?" | If 4+ meetings in agenda: "Heavy meeting day — what was the most important conversation?" If 0-1 meetings: "Lighter meeting day — what did you focus on?" | | 2 | Yes | "Anything unresolved or nagging?" | If action items exist, reference the list: "Any of these feel unresolved or at risk?" | | 3 | If tomorrow's calendar available | "Looking at tomorrow..." | Reference specific meetings: "Tomorrow you have [X] and [Y]. Anything to prep or carry over?" | | 4 | If decisions/wins apparent | "Any decisions worth recording?" | If meeting insights show explicit decisions, reference them. Skip if routine day. | | 5 | Optional | "How was your energy today?" | Only on high-intensity days (4+ meetings or tough signals in notes). |

Wait for the user's responses before continuing.

Step 3: Compose End-of-Day Reflection

Using interview responses, action items, and day context, compose the reflection section:

## End-of-Day Reflection

### Wins
- [Win from interview + artefacts]

### Decisions
- [Decision with [[wikilink]] to entity/project]

### Unresolved / Carry-over
- [Actionable item from interview + unaddressed action items]

### Tomorrow
- [Focus/intent from interview]
- [Prep notes for tomorrow's meetings if mentioned]

### Energy
[One line if answered.]

Rules:

• Apply wikilinks per Guidelines/wikilink-guidelines.md
• 2-5 bullets per subsection
• Omit empty subsections entirely (e.g. if no decisions, omit ### Decisions; if energy not asked or declined, omit ### Energy)
• Keep bullets concise and actionable

Step 4: Idempotency Check and Journal Append

1. Re-read the journal file (fresh read to avoid stale state).
2. If ## End-of-Day Reflection already exists in the file: show the existing content to the user and ask "Replace, append below existing, or skip?"
   • Replace: Remove the existing ## End-of-Day Reflection section and everything below it, then append the new reflection.
   • Append: Add a horizontal rule (---) after the existing reflection, then append the new one.
   • Skip: Do not modify the file. Proceed to Step 5.
3. If ## End-of-Day Reflection does not exist: append the composed section to the end of the journal file.

Step 5: Brag Doc Prompt

If the ### Wins subsection has items: ask the user "Any of these wins worth adding to the brag doc? (y/n/specify which items)".

If yes (or specific items identified), run /brag (read and execute .claude/skills/brag/SKILL.md) with the identified wins as context.

If no or skipped, proceed.

Step 6: Entity and Memory Updates

• Entity stubs: Update any entity stubs with new facts surfaced during the interview (timeline entries, frontmatter changes) per Guidelines/wikilink-guidelines.md stub update rules. Only update if the interview revealed genuinely new information not already captured.
• Session extract: If the interview surfaced substantial new strategic or dynamic context not captured elsewhere, write a session extract to Raw-Materials/Session-Extracts/ following the extract format. Skip for routine responses.

Step 7: Backup

Run /backup (read and execute .claude/skills/backup/SKILL.md) automatically. No confirmation prompt needed.

Step 8: Summary

Display a concise summary:

End-of-day shutdown complete for [Day], [Date]:
- Reflection appended to Curated-Context/Daily-Journals/YYYY-MM-DD.md
- N carry-over items noted in journal [or: No carry-over items]
- N entity stubs updated [or: No entity updates]
- Brag doc updated / skipped
- Changes committed and pushed

If unprocessed queues were flagged in Step 1 and not processed:

Reminder: N items remain in Raw-Materials/

Graceful Degradation

| Missing piece | Behaviour | |---|---| | /slack-fetch fails or returns no messages | Note in summary; continue with existing Slack context from morning | | /slack processing fails | Note in summary; Slack exports remain in Raw-Materials/Slack/ for later | | No daily journal | Warn; offer to create a minimal journal or proceed interview-only | | No meeting insights today | Skip action item consolidation; note "No meeting insights processed today" in interview preamble | | Calendar script fails | Skip tomorrow preview; omit question 3 from interview | | No raw materials to flag | Skip Step 1 entirely | | /backup fails | Report error; reflection is still saved locally | | /brag fails | Report error; wins are still in the journal reflection |