bootstrap

Bootstrap Command

If you see unfamiliar placeholders or need to check which tools are connected, see CONNECTORS.md.

First-run initialization of the productivity system. This is a comprehensive, interactive process that verifies the vault structure, scans all connected sources, and builds a rich understanding of the user's workplace — people, terms, projects, and conventions.

Run this once. After that, workflow skills (configured per-vault) keep everything current.

Workflow Config

The first thing bootstrap does after verifying vault structure is collect the user's team-specific configuration. These values drive all skills:

Key	Description	Example
linear_org	Linear organization slug (from URL)	moov
linear_team	Linear team prefix for ticket IDs	INFRA
team_channels	Slack channels to monitor	#team-infra, #org-infra
github_org	GitHub organization	moovfinancial
standup_note_slug	Filename slug for standup comms notes	infra-standup
review_interval	Days between effort reviews (optional, default: 7)	7

Prompt the user for each value and write them to memory/AGENTS.md under a ## Workflow Config heading:

## Workflow Config
- linear_org: moov
- linear_team: INFRA
- team_channels: #team-infra, #org-infra
- github_org: moovfinancial
- standup_note_slug: infra-standup
- review_interval: 7

Usage

/productivity:bootstrap

Instructions

1. Create Today's Daily Note

Before anything else, ensure today's daily note exists in log/daily/ via Obsidian MCP. This note is where verification tasks and bootstrap findings will be captured. Use the template structure from the task-management skill if creating from scratch.

2. Verify Vault Structure

Check that the expected folders exist in the Obsidian vault via MCP (list_directory or read_note probes). Create any that are missing:

Folder	Purpose
memory/	Agent memory files
memory/context/	Company context
people/	People profiles
efforts/	Multi-step effort tracking
decisions/	Architecture Decision Records
log/daily/	Daily notes
log/cycles/	Linear cycle notes
comms/	Meeting notes, draft communications
+/	Inbox / unorganized notes
z/templates/	Obsidian templates

3. Verify Templates

Check that the required templates exist in z/templates/:

Template	File	Target folder
Daily note	z/templates/periodic-daily.md	log/daily/
Cycle note	z/templates/periodic-cycle.md	log/cycles/
Effort	z/templates/effort.md	efforts/
Decision	z/templates/decision.md	decisions/

If any are missing, warn the user — these templates need to be created before the system works properly.

Templater bindings cannot be verified programmatically (.obsidian/ is inaccessible via MCP). Add a verification task to today's daily note ## Capture section:

- [ ] Verify Templater folder template bindings in Obsidian settings:
    - `log/daily/` → `z/templates/periodic-daily.md`
    - `log/cycles/` → `z/templates/periodic-cycle.md`
    - `efforts/` → `z/templates/effort.md`
    - `decisions/` → `z/templates/decision.md`

4. Check MCP Connections

Before scanning, verify which sources are actually reachable. Try a lightweight call to each:

Source	Test call	Required?
Obsidian	get_vault_stats	Yes — core to the system
Linear	List assigned issues	Recommended
Slack	Search channels	Recommended
GitHub	List PRs	Optional
Gmail	Get profile	Optional
Google Calendar	List calendars	Optional
Notion	Search docs	Optional

Report what's connected and what's not. Skip unavailable sources gracefully — the system works with whatever is connected, and more sources can be added later.

5. Check Existing Memory Content

Read from the Obsidian vault via MCP:

• memory/AGENTS.md — compiled hot cache
• memory/glossary.md — decoder ring
• memory/context/company.md — company context
• people/ — people profiles
• efforts/ — effort tracking

If all of these exist and are populated, this has already been run. Confirm with the user:

It looks like the memory system is already set up:
- X people in people/
- X terms in memory/glossary.md
- X active efforts

Want me to re-scan everything anyway, or is this good?

6. Scan All Connected Sources

Gather data from every available MCP source. Cast a wide net:

Linear:

• Assigned issues (open and recently closed)
• Team members and assignments
• Project documents
• Cycle information

Slack:

• Recent messages in active channels
• DMs (recent, for people and context)
• Active incident channels (#inc- prefix)
• Channel names and purposes

GitHub:

• Open PRs authored by or assigned to user (from github_org in Workflow Config)
• Recent PR reviewers and collaborators

Gmail:

• Recent sent messages (for people and commitments)
• Recent received messages (for context)

Google Calendar:

• Recent and upcoming events (for people, meetings, cadences)
• Recurring meetings (for team processes)

Notion:

• Recently touched long-term docs
• Team wikis and documentation

Obsidian vault:

• Existing daily notes in log/daily/
• Existing efforts in efforts/
• Existing people in people/
• Any existing glossary content

7. Extract and Categorize

From all sources, build lists of:

People — anyone mentioned, messaged, or met with:

• Full name, nickname(s), role, team
• Communication style notes (if observable)
• Frequency of interaction

Terms — shorthand, acronyms, jargon:

• Acronyms and what they expand to
• Internal terms and their meanings
• Project codenames
• Process names

Efforts/Projects — active work:

• Project names and descriptions
• Key people involved
• Current status

Processes — recurring patterns:

• Meeting cadences
• Review processes
• Team rituals

8. Decode Interactively

Present findings grouped by confidence and ask the user to confirm or correct:

## People I Found (23 total)

High confidence — adding these directly:
| Name | Role | Source |
|------|------|--------|
| Mark Alexander | Engineering | Linear, Slack, Calendar |
| Morgan Hargrove | Engineering | Linear, Slack |

Need clarification:
1. **"Steve"** — appears in Slack DMs, 8 mentions. Who is Steve? (full name, role)
2. **"MJ"** — mentioned in Linear comments. Is this a person or abbreviation?

## Terms I Found (15 total)

High confidence:
| Term | Meaning | Source |
|------|---------|--------|
| P0 | Drop everything priority | Linear labels |

Need clarification:
1. **"the migration"** — referenced 12 times in Slack. What project is this?
2. **"ACH"** — is this the standard term or something company-specific?

## Projects/Efforts (8 total)

| Name | Description | Key People |
|------|-------------|------------|
| App migration | Referenced across Linear and Slack | ? |

What am I missing? What did I get wrong?

Work through each category, confirming high-confidence items and asking about unknowns. Skip terms already decoded in earlier rounds.

9. Write Memory Files

From everything gathered, write to the Obsidian vault via MCP:

memory/glossary.md — the full decoder ring:

# Glossary

Workplace shorthand, acronyms, and internal language.

## Acronyms
| Term | Meaning | Context |
|------|---------|---------|

## Internal Terms
| Term | Meaning |
|------|---------|

## Nicknames → Full Names
| Nickname | Person |
|----------|--------|

## Project Codenames
| Codename | Project |
|----------|---------|

memory/context/company.md — tools, teams, processes (use the template from memory-management skill, populated with discovered information).

people/{name}.md — one file per person. Check existing files first — merge new info via patch_note, don't overwrite.

efforts/{name}.md — for active efforts not already tracked. Use the effort naming convention: YYYYMMDDHHMMSS.md, with frontmatter type: effort, status, title, tags: [effort]. See task-management skill for the full schema.

10. Generate memory/AGENTS.md

Compile the hot cache from everything gathered. This is the file agents read first for quick context. See memory-management skill for the exact format.

# Agent Context

> Auto-generated by /productivity:bootstrap. Do not edit manually.
> Source of truth: the rest of this Obsidian vault.
> Last updated: YYYY-MM-DD

## Me
[Name], [Role] on [Team].

## People
(top ~30 by interaction frequency)

## Terms
(~30 most common)

## Active Efforts
(status: active only)

## Preferences
...

11. Report and Establish Working Agreement

Bootstrap complete:
- Vault: structure verified, X folders created
- Templates: all present (Templater binding verification task added)
- Connections: X/7 sources connected
- People: X profiles written (Y new, Z updated)
- Terms: X entries in glossary
- Efforts: X tracked
- AGENTS.md: compiled (X lines)
- Today's note: ready at log/daily/YYYY-MM-DD

Then explain how the system works going forward:

Here's how we'll work together:

**Foundation skills (always available):**
- /productivity:review — triage stale efforts and cycles
- /productivity:bootstrap — re-scan sources and rebuild memory (already done)
- /productivity:start — orient to the system

**Always-on capabilities:**
- Task management — capture, plan, review via daily notes
- Memory — I decode your shorthand, remember your people and projects
- Effort tracking — multi-step work in efforts/ with status lifecycle

**Workflow skills (per-vault):**
Workflow skills like shutdown, sync, and update are configured per-vault to match
your cadence. Check your vault's commands or CLAUDE.md for what's available here.

**What I maintain automatically:**
- memory/AGENTS.md (compiled from vault contents)
- Effort staleness detection (via /productivity:review)

**What you maintain:**
- ## Plan (your intentional work — I never touch this)
- ## Capture (your inbox during the day)
- ## Log (working memory blocks as you work)
- Reviewing and accepting decisions in decisions/
- Templater bindings in Obsidian settings

**Teaching me new things:**
- Say "remember X means Y" anytime
- Correct me when I get something wrong — I'll update the vault
- I'll ask when I hit terms I don't recognize

**Where things live:**
- Daily notes: log/daily/YYYY-MM-DD.md
- Efforts: efforts/YYYYMMDDHHMMSS.md
- Decisions: decisions/DEC-NNNN-slug.md
- People: people/full-name.md
- Memory: memory/AGENTS.md (hot cache), memory/glossary.md (full)

Notes

• Always check existing vault content before writing — merge, don't overwrite
• Present findings for user confirmation before committing to the vault
• High-confidence items can be added directly; uncertain ones must be asked about
• The user's corrections are the most valuable data — pay close attention
• If a source isn't available (MCP not connected), skip it and note the gap
• This is meant to be run once — workflow skills handle ongoing maintenance
• The daily note must exist before adding verification tasks to it (Step 1 handles this)