Brain

Brain

Unified skill for all operations on jonmagic's second brain repository (~/Brain). Covers searching, creating files, capturing content, committing changes, and loading context.

When to Use

Use this skill when the user wants to do anything with the Brain. Common triggers:

| Intent | Example phrases | |--------|----------------| | Search | "search my brain", "what do I know about X", "find notes about X" | | Context | "brain context", "load context for X", "what's the latest on X" | | Commit | "commit brain", "push brain changes", "save my notes" | | Daily project | "new daily project", "create a daily project note for X" | | Bookmark | "bookmark this", "save this link", "bookmark that URL" | | Weekly note | "new weekly note", "start the week", "create this week's note" | | Archive meeting | "archive meeting", "archive my last meeting", "process transcripts" | | Archive session | "archive this session", "save this session to brain" | | Catch me up | "catch me up", "fill in my weekly note", "what happened this week" | | Frontmatter | "add frontmatter", "generate a TID", "backfill frontmatter" |

The user may also say /brain as a convenience.

---

Common Patterns

These patterns are shared across all operations. They are defined here once and referenced throughout.

Git Sync

Before reading or writing any Brain content, sync from remote:

cd ~/Brain && git pull --rebase --autostash

If the pull fails with a merge conflict, abort and report — do NOT auto-resolve:

cd ~/Brain && git rebase --abort

Then tell the user about the conflict so they can resolve it manually. Never use -X theirs or -X ours — these strategies can silently destroy data on frequently-edited files like Weekly Notes.

If the pull fails for a non-conflict reason (network error, etc.), warn and proceed with local state.

Frontmatter & TIDs

All new Brain markdown files require YAML frontmatter:

---
uid:               # Required. 13-char base32-sortable timestamp ID
type:       # Required. See collection types below
created:      # Required. When the content was created
tags: []                # Optional. Controlled vocabulary (see .brain/tag-vocabulary.md)
links:                  # Optional. Structured relationships
  parent: []
  source: []
  related: []
---

Collection types: daily.project, weekly.note, meeting.note, project, snippet, transcript, executive.summary, bookmark, archive, reference

Generate a TID:

ruby ~/.copilot/skills/brain/scripts/generate-tid.rb

Add frontmatter to an existing file:

ruby ~/.copilot/skills/brain/scripts/add-frontmatter.rb 

Options: --dry-run, --type , --created , --force (adds missing fields to existing frontmatter)

Timestamp detection: If the path contains YYYY-MM-DD, that date is used for created. Otherwise, file mtime is used. The script is idempotent — existing frontmatter is not modified unless --force is passed.

Tags: Consult .brain/tag-vocabulary.md before assigning. Pick 1-4 specific tags per file. Tags span six categories: Work Projects, Technical Domains, Activities, Teams, Personal, and Meta.

File Numbering

Many Brain directories use dated folders with sequential numbering:

/YYYY-MM-DD/
  01 first topic.md
  02 second topic.md
  03 third topic.md

When creating files: check the folder for the highest existing number and increment. Create the date folder if it doesn't exist.

Wikilinks

Use Obsidian-style wikilinks to connect documents:

• [[Daily Projects/2026-02-02/05 threadwaste questions evidence.md]]
• [[Projects/hamzo/executive summary]]
• [[uid:TID|Display Text]]

---

Operations

Search & Index

Search and query the Brain's structured index. Supports full-text search, tag filtering, type filtering, link traversal, timeline views, and recent file discovery.

Build the index first (takes ~2 seconds, produces ~/Brain/.brain/index.json):

ruby ~/.copilot/skills/brain/scripts/brain-index.rb ~/Brain --stats

Search modes:

# Full-text search (titles, summaries, paths, tags)
ruby ~/.copilot/skills/brain/scripts/brain-search.rb "proxima abuse"

# Tag search
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --tag hamzo

# Type filter (combinable with tag)
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --type daily.project --tag proxima

# Timeline view (chronological evolution of a topic)
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --timeline "nuanced-enforcement"

# Link traversal (find everything linking to a UID)
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --linked-to 3lz7nwvh4zc2u

# Recent files (modified in last N days)
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --recent 7

# Index statistics
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --stats

Options: --tag , --type , --linked-to , --timeline , --recent , --limit (default: 20), --compact (paths only), --index , --stats

---

Context Loading

Load relevant Brain content at session start. Use when starting work on a project/topic, preparing for a meeting, or answering "what do I know about X?"

Workflow:

1. Sync — cd ~/Brain && git pull --rebase --autostash
2. Rebuild index — ruby ~/.copilot/skills/brain/scripts/brain-index.rb ~/Brain
3. Search — by keyword, tag, type, or timeline depending on what the user needs
4. Read key files in priority order:

• Project executive summary (Projects//executive summary.md)
• Recent Daily Projects on the topic
• Connected files from other time periods

Quick context for a known project:

ruby ~/.copilot/skills/brain/scripts/brain-index.rb ~/Brain
cat ~/Brain/Projects//executive\ summary.md
ruby ~/.copilot/skills/brain/scripts/brain-search.rb --tag  --type daily.project --limit 5

---

Commit & Push

Analyze Brain changes, create semantic commits, and push. Non-interactive — do the best job without asking for confirmation.

Workflow:

1. cd ~/Brain
2. Sync from remote (see Common Patterns)
3. git status to see changes
4. Group files by type/purpose and create commits:

| Directory | Commit format | |-----------|--------------| | Weekly Notes/ | docs(weekly): ... | | Daily Projects/ | docs(daily): ... | | Meeting Notes/ | docs(meetings): ... | | Snippets/ | docs(snippets): ... | | Executive Summaries/ | docs(summaries): ... | | Transcripts/ | docs(transcripts): ... | | Projects/ | docs(projects): ... | | Bookmarks/ | docs(bookmarks): ... | | Archive/ | chore(archive): ... | | Skills/config | chore: ... |

1. git push after all commits
2. Report what was committed

Safety rules:

• Never auto-resolve merge conflicts
• Never force push
• Never amend commits unless explicitly asked
• Show what will be committed but do not pause for confirmation

Auto-commit cron (scripts/auto-commit.sh): Hourly safety net that commits and pushes uncommitted changes. Install with crontab -e:

0 * * * * ~/.copilot/skills/brain/scripts/auto-commit.sh

---

Create Daily Project Note

Create a numbered markdown file in today's Daily Projects/YYYY-MM-DD/ folder.

Workflow:

1. Sync from remote
2. User provides a title/description
3. Create Daily Projects/YYYY-MM-DD/ folder if needed
4. Find next available number prefix (01, 02, ...)
5. Create NN title.md with frontmatter:

---
uid: 
type: daily.project
created: T00:00:00Z
session: 
---

# title

Session auto-detection for the session frontmatter field:

| Tool | Env Variable | Resume Command | |------|-------------|----------------| | OpenCode | OPENCODE=1 | opencode -s | | Copilot | COPILOT_PROXY_TOKEN_CMD | copilot --resume= | | Claude | CLAUDE_CODE=1 | claude --resume |

Run ruby ~/.copilot/skills/brain/scripts/detect-session.rb --project-dir ~/Brain to detect. If it exits non-zero, omit the session field.

---

Create Bookmark

Save a URL as a dated bookmark file with structured frontmatter and an executive-summary-style narrative body.

File format: Bookmarks/YYYY-MM-DD/NN title.md

---
uid: 
type: bookmark
created: T00:00:00Z
url: https://example.com/article
title: "Article Title"
source: hacker-news
tags: [pkm, architecture]
---

Workflow:

1. Sync from remote
2. Extract URL(s) from user message
3. Get optional context: blurb, source (e.g. hacker-news, slack), tags
4. Fetch page title via WebFetch
5. Generate body: user's blurb if provided, otherwise generate executive-summary-style narrative
6. Create file with frontmatter + body
7. Confirm to user

Body rules (when generating, not when user provides blurb):

1. Start with a clear title line (not a markdown header)
2. Narrative paragraphs only — no bullets, headers, or lists
3. Contextual linking: link to source when referencing claims. For GitHub: @username unlinked, with separate ([ref](URL)) links
4. Focus on critical content: key arguments, evidence, implications, business impact
5. Close with personal relevance paragraph
6. Formal tone, dense prose

If fetch fails or returns thin metadata, still create the bookmark with a short narrative based on title/URL.

Multiple URLs: Create one file per URL with sequential numbering.

---

Create Weekly Note

Create a weekly planning note from template with dates filled in, then interactively review carryover items from last week.

Workflow:

1. Sync from remote
2. Run: ~/.copilot/skills/brain/scripts/create-weekly-note
3. Review previous week's Active Projects — confirm what's still active
4. Review unchecked items from previous week's Parked, Sessions, TODO sections:

• Mark done, keep for this week, or drop

1. Copy confirmed carryover items to new note
2. Add frontmatter:

---
uid: 
type: weekly.note
created: 
tags: []
links:
  related: []
---

Template: Edit assets/weekly-note-template.md to customize. Placeholders: {{sunday:YYYY-MM-DD}} through {{saturday:YYYY-MM-DD}}. The default recurring schedule includes {{Meeting Notes/tgthorley}} on Tuesdays at 1400.

---

Archive Meeting

Archive meeting transcripts (Zoom folders or Teams .vtt files) into structured Brain artifacts.

Outputs:

• Transcripts/YYYY-MM-DD/NN.md — imported transcript
• Executive Summaries/YYYY-MM-DD/NN.md — generated summary
• Meeting Notes//YYYY-MM-DD/NN.md — meeting notes
• Weekly note checkbox marked complete

Workflow — use the CLI:

archive-meeting \
    --brain-dir ~/Brain \
    --executive-summary-prompt-path ~/code/jonmagic/prompts/summarize/zoom-transcript-executive-summary.md \
    --detailed-notes-prompt-path ~/code/jonmagic/prompts/summarize/transcript-meeting-notes.md

Interactive mode uses fzf to select from recent meetings and pending meeting note targets.

Explicit mode (when user specifies input and target):

archive-meeting \
    --brain-dir ~/Brain \
    --input "/path/to/transcript.vtt" \
    --meeting-notes-target "jonmagic" \
    --executive-summary-prompt-path ~/code/jonmagic/prompts/summarize/zoom-transcript-executive-summary.md \
    --detailed-notes-prompt-path ~/code/jonmagic/prompts/summarize/transcript-meeting-notes.md

CLI options: --brain-dir, --input, --meeting-notes-target, --executive-summary-prompt-path, --detailed-notes-prompt-path, --date, --dry-run

Prerequisites: The archive-meeting CLI from ~/code/jonmagic/scripts must be on PATH. Verify with which archive-meeting.

Frontmatter for outputs: Add to any files the CLI doesn't frontmatter automatically. Link executive summaries and meeting notes back to their transcript via links.source: [].

References: references/formats.md (Zoom/Teams conventions), references/meeting_notes_routing.yml (routing overrides)

---

Archive Session

Archive a Copilot CLI session to the Brain — creates a daily project file with the session transcript and adds a resume link to the weekly note.

Inputs needed: Session ID (from session context path ~/.copilot/session-state/UUID/) and focus (1-5 word description from user).

Workflow:

1. Sync from remote
2. Ask user for the focus
3. Extract transcript:

``bash ~/.copilot/skills/brain/scripts/extract-transcript \ --session-id "SESSION_ID" \ --output "/tmp/session-transcript.md" ``

1. Create daily project file:

``bash ~/.copilot/skills/brain/scripts/create-daily-project \ --focus "FOCUS" \ --session-id "SESSION_ID" \ --transcript "/tmp/session-transcript.md" ``

1. Add frontmatter to created file:

``bash ruby ~/.copilot/skills/brain/scripts/add-frontmatter.rb "PATH_FROM_STEP_4" ``

1. Update weekly note:

``bash ~/.copilot/skills/brain/scripts/update-weekly-note \ --focus "FOCUS" \ --session-id "SESSION_ID" \ --daily-project-path "PATH_FROM_STEP_4" ``

1. Report the daily project path and which weekly note was updated

---

Catch Me Up

Synthesize daily activity (daily projects, meeting notes, GitHub issues/PRs) into narrative summaries for weekly note day sections. Fills gaps where the user didn't write context notes.

Workflow:

1. Sync from remote
2. Determine date range:

• "this week" → current week (Sunday through today)
• "last week" → previous full week
• "catch me up" with no qualifier → current week

1. Read the weekly note — identify which day sections need filling:

• A day has substantive content only if it has narrative bullets (sentences describing what happened)
• Just wikilinks, session commands, or - placeholders do NOT count

1. Gather activity:

``bash ~/.copilot/skills/brain/scripts/gather-activity --week YYYY-MM-DD --lines 100 ` Options: --date, --from/--to, --files (paths only), --no-github, --github-user `

1. Read full content of daily projects and meeting notes for days needing summaries
2. Write first-person narrative bullets under each day heading

GitHub activity classification — critical rules:

The gather-activity script classifies GitHub items into summary-safe and context-only:

• issues / pull_requests arrays → safe to summarize (created or merged that day)
• context_only_issues / context_only_pull_requests → advisory only, ignore unless corroborated by daily project or meeting note

Activity kind rules:

| activity_kind | Safe to say | |-----------------|-------------| | opened | User opened/filed it that day | | merged | User merged/shipped it that day | | opened_and_merged | Opened and merged same day | | updated_existing | NOT safe for day-specific claims — skip unless corroborated |

Validation gate: Never write "opened", "filed", "merged", "shipped", "reviewed" or similar day-specific language unless backed by a summary-safe GitHub item or corroborating daily project / meeting note. Fail closed on ambiguity.

Voice and format:

• First person, informal, concise
• Short bullet points (one line each)
• Capture what was worked on, key decisions, what happened
• Include meeting highlights: who, key outcomes
• Include session resume commands from daily project frontmatter
• Link to daily project files with wikilinks
• Tone: quick capture of state of mind and progress, like end-of-day notes

Good example:

## Monday
- rescheduled with znull to next week to talk about the abuse data plane idea
- attended team sync
- synced with romanofoti
- started the morning by getting spamurai-next and hamzo PRs closer to shipping

Avoid: listing just file/meeting names, corporate language, bold prefixes, fabricating information.

Day heading formats: Day section headings may use either ## Monday or ## Monday (2026-02-09). Match whichever format the existing note uses.

Options: Augment existin

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: jonmagic
• Source: jonmagic/skills
• License: ISC

Install and usage instructions live in the source repository linked above.