dcurlewis/.claude/skills/setup

Setup

Interactive setup wizard that configures the AI Context System for the user's specific role, organisation, and toolchain. Safe to re-run; it should refresh the user's choices without disturbing unrelated content.

Process

Work through these phases in order. At each phase, ask the user questions and update files based on their responses. If the user wants to skip a phase, respect that and move on.

Phase 1: Identity & Role

Check if CLAUDE.md still contains <YOUR_NAME_HERE> placeholders.

If placeholders exist, ask:

1. What is your name?
2. What is your job title or role?
3. What company or organisation do you work for?
4. What city or timezone are you in?
5. How should the AI think of its role? Suggest options based on their title:
   • Engineering Manager / Director / VP → "a senior engineering leadership coach"
   • Software Engineer / IC → "a technical mentor and productivity partner"
   • Product Manager → "a strategic product advisor"
   • Recruiter / Talent → "a hiring strategy partner"
   • Designer → "a design thinking and collaboration partner"
   • Other → "a professional knowledge management assistant"
   • Or: "describe it in your own words"

Replace placeholders in CLAUDE.md:

• <YOUR_NAME_HERE> → their name
• <YOUR_ROLE> → their title
• <YOUR_COMPANY> → their company
• <AI_ROLE_DESCRIPTION> → chosen AI role framing
• <YOUR_CITY_OR_TIMEZONE> → their location

Also replace <YOUR_NAME_HERE> in Curated-Context/System/about-me.md.

If placeholders are already filled, confirm the current values and ask if they want to update anything.

Phase 2: Profile Interview

Walk through the sections of Curated-Context/System/about-me.md conversationally:

1. Professional identity, role, team structure, reporting line
2. Scope and responsibilities
3. Collaboration context, key people they work with
4. Core values and leadership style
5. Strengths and growth areas
6. How AI should best support them

Write their responses into the appropriate sections. Keep the tone professional but natural. It is fine to leave sections sparse; they will fill in over time.

Phase 3: Entity Types

Based on their role, suggest entity types. People is always included. Suggest 3-4 additional types:

| Role | Suggested entity types | | ---- | ---------------------- | | Engineering Manager | People, Teams, Projects, Vendors | | Software Engineer | People, Projects | | Product Manager | People, Products, Projects | | Recruiter | People, Candidates, Roles, Agencies | | Designer | People, Projects, Clients, Design Systems | | Data Scientist | People, Projects, Datasets, Models | | Generic | People, Projects |

Present the suggestions and let the user adjust: "Based on your role as [X], I'd suggest tracking these entity types: [list]. Would you like to add, remove, or rename any?"

For each confirmed entity type:

1. Create the subdirectory under Curated-Context/Entities/
2. Add a .gitkeep file
3. Update the ## Entity Types section in CLAUDE.md with the type name, directory, description, and suggested frontmatter fields
4. Update the ### Stub Directories table in Guidelines/wikilink-guidelines.md

Phase 4: Team & Org Seeding

Ask the user about their immediate team and organisation:

1. Who are your direct reports or closest collaborators? (names, roles)
2. What teams or groups do you work with?
3. Any key acronyms or terminology at your company?

For each person mentioned:

• Create a minimal entity stub in Curated-Context/Entities/People/
• Add to the Canonical Names list in Guidelines/wikilink-guidelines.md

For each team mentioned (if Teams is a configured entity type):

• Create a minimal team stub

Update Curated-Context/System/reference.md with:

• acronyms and abbreviations
• organisational structure (if provided)
• company terminology

Phase 5: Memory Seeding

Ask about current priorities: "What are the 3-5 most important things you're working on or tracking right now?"

Write topic-appropriate content into:

• Memory/memory-people.md — key relationships and dynamics
• Memory/memory-work.md — active projects and deliverables
• Memory/memory-context.md — strategic context, vendor relationships, budget

Update Curated-Context/System/memory-index.md:

• set the current priorities list
• update the last-updated date

Phase 6: Integrations (Optional)

Ask: "What external tools do you use? (Jira, Linear, GitHub, GitLab, Slack, Google Calendar, Notion, etc.)"

For each tool:

1. Check if an MCP server is already available
2. If available, note it and explain how the system can use it
3. If not available, point to Guidelines/data-sources.md for setup instructions
4. For tools with reference connectors (Jira, GitHub, Calendar, RSS), offer to configure Scripts/sync_config.json

Update Scripts/sync_config.json with their configuration when they want that help.

Phase 7: Morning Routine Design

This phase personalises .claude/skills/morning/SKILL.md.

7a. Explain the model

Read:

• .claude/skills/morning/SKILL.md
• .claude/skills/setup/morning-modules/README.md

Then explain briefly:

• /morning always keeps a useful generic baseline
• optional additions are compiled into the generated section by /setup
• rerunning /setup replaces the previous generated section with the new choices

7b. Interview the user about optional morning behaviour

Do this conversationally. Do not dump a giant questionnaire all at once. Offer examples, then ask only the follow-up questions needed for the options they choose.

Suggested examples to offer:

• Slack summaries or communication highlights
• recurring weekday routines like news, delivery review, or meeting analysis
• task-board sync
• graph or Obsidian-derived context
• bespoke recurring prompts, reminders, or checks

Ask about these categories:

1. Communication context
   • Do they want Slack summaries or similar highlights included when available?
2. Recurring weekday routines
   • Do they want /news, /delivery, or /meeting-analysis launched from /morning?
   • If yes, on which weekday should each run?
3. Task management
   • Do they want /morning to read from and update Curated-Context/System/My Tasks.md?
4. Knowledge hygiene
   • Do they want graph or Obsidian-derived context in the morning briefing?
   • If yes, on which weekday should the deeper relationship digest appear?
5. Bespoke custom routines
   • Ask: "Are there any recurring prompts, reminders, or checks you want /morning to include in your own words?"

If the user declines all optional behaviour, keep the default placeholder text in the generated section.

7c. Build the generated module section

Read only the snippet files needed from .claude/skills/setup/morning-modules/.

Available reusable snippets:

• slack-highlights.md
• weekly-news.md
• delivery-review.md
• meeting-analysis.md
• task-board-sync.md
• graph-intelligence.md
• custom-routine-template.md

Assemble the generated section using the deterministic order documented in README.md.

For each selected snippet:

1. Replace any placeholders such as:
   • <NEWS_DAY_NAME>
   • <DELIVERY_DAY_NAME>
   • <MEETING_ANALYSIS_DAY_NAME>
   • <GRAPH_REVIEW_DAY_NAME>
2. Keep the snippet heading and hook labels intact
3. Trim examples or options that the user did not choose

For bespoke routines:

1. Read custom-routine-template.md
2. Translate the user's request into one or more small module blocks using the same hook structure
3. Keep custom modules concise and practical
4. Append them after the reusable modules

7d. Replace only the generated section in /morning

Update .claude/skills/morning/SKILL.md by replacing only the content between:

• <!-- GENERATED: morning-modules:start -->
• <!-- GENERATED: morning-modules:end -->

Rules:

• Do not rewrite the stable baseline outside those markers
• On rerun, fully replace the previous generated content
• If the user selected no optional modules, restore the default placeholder text
• Preserve the explanatory text and the markers themselves
• Keep the generated content readable in diffs, with one module block after another in deterministic order

Phase 8: Knowledge Base Import (Optional)

Ask: "Do you have an existing knowledge base, notes folder, or documents you'd like to import? (The import is read-only; your original files won't be modified.)"

If yes, run /import (read and execute .claude/skills/import/SKILL.md).

Phase 9: Generate Dashboards

Based on the configured entity types, update the Dataview queries in Curated-Context/Dashboards/:

• People-Dashboard.md — always relevant
• Active-Projects.md — if Projects or similar entity type exists
• Team-Overview.md — if Teams entity type exists
• Stale-Relationships.md — always relevant

Remove or hide dashboards for entity types that were not configured.

Phase 10: Summary & Next Steps

Display a summary of everything configured:

Setup complete. Here's what was configured:

Identity: [Name], [Role] at [Company]
Entity types: [list]
People stubs created: N
Team stubs created: N
Memory files seeded: Yes/No
Integrations: [list of configured tools]
Morning modules: [selected modules or "Baseline only"]

Next steps:
1. Drop a meeting transcript into Raw-Materials/Meeting-Transcripts/ and run /meeting
2. Run /morning each day for your daily briefing
3. Run /help to see all available commands
4. After a week of use, run the memory update cycle to consolidate context

Notes

• This command is conversational and interactive. Ask questions, wait for responses, then act.
• Keep the conversation moving; do not spend too long on any single phase.
• It is fine to skip phases the user is not ready for. They can re-run /setup later.
• All file writes should be additive and respectful of existing content, except the generated section in .claude/skills/morning/SKILL.md, which is intentionally replaced on rerun.