Wiki Tags Refresh

wiki-tags-refresh

Read ~/.copilot/installed-plugins/ihudak-copilot-plugins/obsidian-llm-wiki/skills/wiki-schema/SKILL.md fully before proceeding.

Invoke with: /wiki-tags-refresh [directory]

If no directory argument is given, scan the entire vault (excludes .obsidian/).

Step 1 — Resolve vault path

Set VAULT using the recommended fallback pattern:

VAULT="${VAULT_PATH:-${VAULT:-${HOME}/obsidian_vault}}"

VAULT_PATH takes precedence; VAULT is accepted as an alias; ~/obsidian_vault is the default. WSL users with a Windows-side vault must set VAULT_PATH explicitly (e.g. /mnt/c/Users/<name>/obsidian_vault).

Step 2 — Read the current tag index

Read .obsidian/copilot/tag-index.md fully. Extract every documented tag (lines matching `#tagname`).

Step 3 — Collect tags from pages

Resolve the scan root:

SCAN_ROOT="${VAULT_PATH:-${VAULT:-${HOME}/obsidian_vault}}"

If /wiki-tags-refresh was invoked with a directory argument, append it instead: scan every .md file under the vault root by default (excludes .obsidian/), or under "${SCAN_ROOT}/<directory>" if a directory argument was given.

The frontmatter extraction (YAML multi-line) across $SCAN_ROOT:

find "$SCAN_ROOT" -name "*.md" -not -path "*/.obsidian/*" -print0 | xargs -0 awk '
  FNR==1{front=0; intags=0}
  /^---/{if(front==0) front=1; else front=0; next}
  !front{next}
  /^tags:/{intags=1; next}
  /^[a-zA-Z]/{intags=0; next}
  intags && /^  - /{gsub(/^  - /, ""); print}
' | sort -u

Note: only the YAML multi-line block form (tags: followed by - tagname lines) is parsed; inline arrays (tags: [a, b]) are silently skipped by this script. Worth knowing now that the scan covers the whole vault, where inline-array frontmatter is common outside plugin-authored wiki pages. -print0/xargs -0 is required — vault filenames routinely contain spaces (e.g. Daily/2026-07-13 Standup.md), which would otherwise be word-split into separate, unopenable arguments.

Also collect inline tags from page bodies across the same scope:

grep -roh --exclude-dir=".obsidian" "#[a-zA-Z][a-zA-Z0-9/_-]*" "$SCAN_ROOT" | sort -u

Exclude false positives by inspecting each match's source line: the tag pattern already requires a letter immediately after #, so standard # Heading / ## Heading lines never match it. The main residual false-positive source is URL fragments (e.g. page.md#Section) and code blocks — use judgment when presenting candidates in Step 6.

Deduplicate the full collected set.

This walks the vault tree twice (once for frontmatter, once for inline tags); on a large vault expect it to take longer than the previous wiki/-only scan.

Step 4 — Diff against tag-index.md

• New tags: found in scanned pages (default: entire vault) but NOT documented in tag-index.md
• Stale candidates: documented in tag-index.md but found in ZERO scanned pages (flag only — never auto-remove; they may be used in other vault files)

Step 5 — Report findings

wiki-tags-refresh results
─────────────────────────
New (undocumented in tag-index.md):   #newtag  #anothertag
Stale (0 uses in scanned pages):      #oldtag  #unusedtag
Already in sync:                      N tags

Step 6 — Process new tags

For each new tag, ask the user:

1. "What does #newtag mean?"
2. "Which section of tag-index.md should it go in?" Present the existing section list as choices + "Other / new section…"
3. If the user skips a tag, leave it undocumented.

Step 7 — Process stale candidates

Present the stale list and ask:

• "Remove all stale tags from index" — removes only confirmed zero-use tags
• "Review one by one" — step through each with keep/remove choice
• "Keep all (ignore)" — leave index unchanged for these

Note: Step 3's scan may have been directory-scoped (if a directory argument was given to /wiki-tags-refresh), so a "stale" candidate might still be in use elsewhere in the vault. Always confirm zero vault-wide use before suggesting removal, using an exact tag match — a substring match like grep "#tagname" would false-match #tagname-extended or #tagname/nested — and excluding only tag-index.md itself: every documented tag appears in tag-index.md by definition, so excluding it (not the whole .obsidian/ directory) avoids a false "always used" result while still letting genuine uses elsewhere in .obsidian/ (e.g. custom prompts) protect a tag from removal:

grep -roh --include="*.md" --exclude="tag-index.md" "#[a-zA-Z][a-zA-Z0-9/_-]*" "${VAULT_PATH:-${VAULT:-${HOME}/obsidian_vault}}" \
  | grep -xF "#tagname" | wc -l

If vault-wide count > 0, flag as "used elsewhere in the vault — keep?" rather than suggesting removal.

Step 8 — Update tag-index.md

For approved new tags:

• Insert into the chosen section in alphabetical order within the section.
• Format: `#tagname` - <description>
• Before appending to ## Recently Approved Tags, ensure the section exists:

  grep -q "^## Recently Approved Tags" ".obsidian/copilot/tag-index.md" || \
    printf "\n## Recently Approved Tags\n" >> ".obsidian/copilot/tag-index.md"
  

• Append to ## Recently Approved Tags: `#tagname` (YYYY-MM-DD) - <description>

For confirmed removals:

• Remove the entry from its section.
• Do NOT remove from ## Recently Approved Tags (history record).

Do not touch any section not affected by the diff.

Step 9 — Append to _log.md

## [YYYY-MM-DD] tags-refresh | N new tags proposed, M approved, L removed
- New: #tag1, #tag2 (if any approved)
- Removed: #oldtag (if any removed)

Finish

wiki-tags-refresh complete.
Tags added: N   Tags removed: M   Unchanged: K
tag-index.md updated.