vault

name	vault
description	Read and write notes in the Obsidian vault. Use for task logs, knowledge capture, and building context.

Vault (Obsidian Notes)

Location: ~/Documents/Notes/ (vault name: Notes)

Principles

1. Obsidian vault for documents — detailed notes, task logs, project context
2. Plain Unix commands for vault operations — treat the vault as Markdown files under ~/Documents/Notes; do not use the obsidian CLI
3. WikiLinks for connections — build traversable knowledge graph
4. ALWAYS: New tasks should be in an open state

Templates

Templates are in the skill directory, not the notes directory:

• .codex/skills/vault/templates/knowledge-note.md
• .codex/skills/vault/templates/task.md
• .codex/skills/vault/templates/recipe.md
• .codex/skills/vault/templates/weekly-summary.md

Unix Commands

Set a vault variable when commands get long:

VAULT="$HOME/Documents/Notes"

Spaces in paths: quote the whole path:

sed -n '1,220p' "$HOME/Documents/Notes/Projects/2025-10 Patchwork/Note.md"

Creating and writing

mkdir -p "$VAULT/<folder>"
printf '%s\n' "<text>" > "$VAULT/<folder/file.md>"
cp "$HOME/.codex/skills/vault/templates/<template>.md" "$VAULT/<folder/file.md>"
printf '%s\n' "<text>" >> "$VAULT/<folder/file.md>"
tmp="$(mktemp)" && { printf '%s\n' "<text>"; cat "$VAULT/<folder/file.md>"; } > "$tmp" && mv "$tmp" "$VAULT/<folder/file.md>"

Reading

sed -n '1,220p' "$VAULT/<folder/file.md>"
cat "$VAULT/<folder/file.md>"
stat "$VAULT/<folder/file.md>"
rg -n '^#{1,6} ' "$VAULT/<folder/file.md>"

Searching and finding

rg -n "<text>" "$VAULT/<folder>"
find "$VAULT/<folder>" -type f -name '*.md' | sort
rg -n "(^|[[:space:]])#<tag>\\b" "$VAULT"
rg -oh "(^|[[:space:]])#[A-Za-z0-9/_-]+" "$VAULT" | sed 's/^[[:space:]]*//' | sort | uniq -c | sort -nr

Properties (frontmatter)

awk '/^---$/ {block++; next} block == 1 && /^<prop>:/ {print}' "$VAULT/<folder/file.md>"
perl -0pi -e 's/(^---\n(?:.*\n)*?)^status:.*$/${1}status: done/m' "$VAULT/<folder/file.md>"
perl -0pi -e 's/^<prop>:.*\n//m' "$VAULT/<folder/file.md>"
awk '/^---$/ {block++; next} block == 1 {print} block == 2 {exit}' "$VAULT/<folder/file.md>"

Graph queries (linking)

note="$(basename "$VAULT/<folder/file.md>" .md)"; rg -n "\\[\\[$note(\\||\\])" "$VAULT"
rg -o "\\[\\[[^]|#]+(#[^]|]+)?(\\|[^]]+)?\\]\\]" "$VAULT/<folder/file.md>" | sort -u
for f in $(find "$VAULT" -type f -name '*.md'); do name="$(basename "$f" .md)"; rg -q "\\[\\[$name(\\||\\])" "$VAULT" || printf '%s\n' "$f"; done
rg -L "\\[\\[[^]]+\\]\\]" "$VAULT" -g '*.md'
rg -o "\\[\\[[^]|#]+(#[^]|]+)?(\\|[^]]+)?\\]\\]" "$VAULT" | sed -E 's/.*\\[\\[([^]|#]+).*/\\1/' | sort -u

Tasks

rg -n '^\\s*- \\[ \\] ' "$VAULT"
rg -n '^\\s*- \\[x\\] ' "$VAULT"
rg -n '^\\s*- \\[[ x]\\] ' "$VAULT/<folder/file.md>"
perl -0pi -e 's/^(\s*- )\[ \]/$1[x]/m' "$VAULT/<folder/file.md>"

File management

mkdir -p "$(dirname "$VAULT/<new-path>")"
mv "$VAULT/<old-path>" "$VAULT/<new-path>"
rm "$VAULT/<folder/file.md>"

Time-based file queries

find "$VAULT" -type f -name '*.md' -mtime -7 | sort
find "$VAULT/Projects" -path '*/Tasks/*.md' -type f -mtime -7 | sort

Timestamps

Always use real timestamps, never placeholders:

date +"%Y-%m-%d %H%M%S"    # task filename
date +"%Y-%m-%d %H:%M"      # log entry header
date -Iseconds               # frontmatter (ISO-8601)

Task File Path

~/Documents/Notes/Projects/<YYYY[-MM] Project>/Tasks/<YYYY-MM-DD HHMMSS> <Title>.md

Linking Strategy

Link if it improves the note, not just because it matches a term.

1. Semantic discovery — rg -n "<concept>" "$VAULT"
2. Backlinks — note="$(basename "$file" .md)"; rg -n "\\[\\[$note(\\||\\])" "$VAULT"
3. Tags overlap — rg -n "(^|[[:space:]])#<tag>\\b" "$VAULT"
4. Add as WikiLinks using breadcrumb pattern: [[Parent]] | [[Related]]

Capture Heuristics

Worth capturing when: principle applies across contexts, caused debugging time, method that saves time later, non-obvious choice with reasoning worth preserving, link to documentation.

Where to capture:

Destination	When
Existing note	Extends/refines an existing topic (search first)
New note	Substantial, standalone, referenceable
Task log only	One-off detail that won't generalise
Project recipe	Repeatable steps specific to this project

Note Locations

Folder	Purpose
Development/	Conceptual topics, paradigms, architectural patterns
HowTo/	Procedural guides, techniques, not project-specific
Tools/	Software tools and their usage
Projects/<project>/	Project-level knowledge
Projects/<project>/Tasks/	Task logs only — never knowledge notes
Projects/<project>/Glossary/	Glossary entries
Projects/<project>/Recipes/	Project-specific recipes and runbooks
Projects/<project>/Meetings/	Meetings with date prefix
Journal/Weekly Notes/	Weekly summaries (generated from task activity)

Extraction signal: When a task log contains repeatable steps, extract into a recipe.

Task Reference Resolution

Every mention of a JIRA issue number must be a wiki-link to its task note. Never leave bare issue numbers.

1. Build lookup from find results: issue number → full filename
2. If not found: rg -n "RH-6949" "$VAULT/Projects"
3. Use aliased wiki-links: [[2026-02-13 141534 RH-6949 Performance issue|RH-6949]]

Applies to all sections — summaries, blockers, carryover, etc.

Generating Weekly Summaries

Path: ~/Documents/Notes/Journal/Weekly Notes/<YYYY>-W<WW>.md

Data gathering

find "$VAULT/Projects" -path '*/Tasks/*.md' -type f -mtime -7 | sort
rg -l '^status: in-progress$|^status: "in-progress"$' "$VAULT/Projects"
find "$VAULT/Projects" -type f -name '*.md' -mtime -7 ! -path '*/Tasks/*' | sort
find "$VAULT/Projects" -path '*/Recipes/*.md' -type f -mtime -7 | sort

Workflow

1. Gather modified tasks, new notes, outstanding tasks from previous week
2. Read each: sed -n '1,220p' "$task"
3. Build JIRA → filename lookup (see Task Reference Resolution)
4. Populate weekly summary template, wiki-linking all references
5. Create: printf '%s\n' "<text>" > "$VAULT/Journal/Weekly Notes/<YYYY>-W<WW>.md"

Generating Daily Notes

Path: ~/Documents/Notes/Journal/Daily Notes/<YYYY>/<YYYY-MM>/<YYYY-MM-DD>.md

Data gathering

find "$VAULT/Projects" -path '*/Tasks/*.md' -type f -mtime -1 | sort
find "$VAULT/Projects" -path "*/Meetings/$(date +%Y-%m-%d)*.md" -type f | sort
find "$VAULT/Projects" -type f -name '*.md' -mtime -1 ! -path '*/Tasks/*' ! -path '*/Meetings/*' | sort

Slack conversations

Search Slack for important conversations the user is involved in today. Use slack_search_public_and_private with from:<@USER_ID> and date filters to find messages the user sent. Then read the threads of those messages to get full context. Only include conversations the user participated in. Exclude DMs.

For each noteworthy conversation, capture:

• A one-line summary
• A link to the thread (from the message permalink)

Format

---
---
# [[YYYY-MM-DD]]

## Work
- [[Task link|Short name]] — one-line summary

## Meetings
- [[Meeting note]]

## Notes created
- [[Note name]] (type)

## Slack conversations
- One-line summary of discussion — [link](permalink)

Workflow

1. Find tasks, meetings, notes modified today
2. Read each: sed -n '1,220p' "$task"
3. Search Slack for conversations the user participated in today (from:<@USER_ID>, no DMs), then read threads for context
4. Build JIRA → filename lookup (see Task Reference Resolution)
5. Create or append daily note, wiki-linking all references

What NOT to include

• DO NOT include changed files