Skill

update-vault

Refresh the vault — pull plugin updates and regenerate the skill-index used by vault-search. Use when the user asks to "update the vault" or after adding new plugins to the inventory.

npx claudepluginhub danielrosehill/claude-code-plugins --plugin claude-vault

Tool Access

This skill uses the workspace's default tool permissions.

Preview

Two jobs: pull updates for vault-tracked plugin sources, and regenerate `skill-index.txt` so `vault-search` reflects current state.

SKILL.md

Similar Skills

cache-components

139.3k

Guides Next.js Cache Components and Partial Prerendering (PPR): 'use cache' directives, cacheLife(), cacheTag(), revalidateTag() for caching, invalidation, static/dynamic optimization. Auto-activates on cacheComponents: true.

cache-components

mcp-builder

124.2k

Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).

9 files

anthropics-skills-13

Stats

Stars0

Forks0

Last CommitApr 30, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

update-vault

Two jobs: pull updates for vault-tracked plugin sources, and regenerate skill-index.txt so vault-search reflects current state.

Steps

Resolve the inventory and vault dir per skills/_shared/inventory-path.md. Read inventory.json. If missing, tell the user to create it (per README) and stop.

Plugin updates (best-effort):

For each plugin entry, look for the marketplace cache directory (typically ~/.claude/plugins/cache/<marketplace>/<plugin-name> or similar). If it's a git checkout, run git -C <path> pull --ff-only and capture the result (updated / already up to date / failed).
If the path isn't a git repo or doesn't exist, mark as skipped (not git) — don't fail the whole run.
Collect results into a per-plugin status line.

MCP servers — surface available updates without applying:

For npm-style entries (command: npx, args containing -y <pkg>), note the package name. Don't auto-update; just report.
Skip everything else.

Skill-index regeneration (per skills/_shared/skill-index.md):

Open $VAULT_DIR/skill-index.txt for writing (truncate).
For each plugin in the inventory:
- If the entry has a skills array ([{name, description}]), emit one line per skill: <name> | <plugin> | <description> (strip newlines and pipes from description, trim to ~100 chars).
- Else, attempt best-effort discovery: scan the plugin's installed directory for skills/*/SKILL.md files; parse the frontmatter name and description; emit those lines. (Use head -20 per file plus a simple awk/grep for the frontmatter — don't load full bodies.)
- If neither yields anything, emit a single placeholder: <plugin-name> | <plugin-name> | (no skill metadata — see plugin README).
Sort the output for stable diffs.

Report a summary: N plugins pulled (X updated, Y already current, Z skipped), M skill-index rows written.

Safety

Never run git pull outside the marketplace cache directory.

Never modify the inventory file itself — this skill is read-only against inventory.json.

If a plugin directory has uncommitted local changes, skip the pull and surface the warning.

Output style

One status line per plugin during the pull phase, then a totals line, then the index regeneration totals. Don't dump file contents.

update-vault

Two jobs: pull updates for vault-tracked plugin sources, and regenerate skill-index.txt so vault-search reflects current state.

Steps

Resolve the inventory and vault dir per skills/_shared/inventory-path.md. Read inventory.json. If missing, tell the user to create it (per README) and stop.

Plugin updates (best-effort):

For each plugin entry, look for the marketplace cache directory (typically ~/.claude/plugins/cache/<marketplace>/<plugin-name> or similar). If it's a git checkout, run git -C <path> pull --ff-only and capture the result (updated / already up to date / failed).
If the path isn't a git repo or doesn't exist, mark as skipped (not git) — don't fail the whole run.
Collect results into a per-plugin status line.

MCP servers — surface available updates without applying:

For npm-style entries (command: npx, args containing -y <pkg>), note the package name. Don't auto-update; just report.
Skip everything else.

Skill-index regeneration (per skills/_shared/skill-index.md):

Open $VAULT_DIR/skill-index.txt for writing (truncate).
For each plugin in the inventory:
- If the entry has a skills array ([{name, description}]), emit one line per skill: <name> | <plugin> | <description> (strip newlines and pipes from description, trim to ~100 chars).
- Else, attempt best-effort discovery: scan the plugin's installed directory for skills/*/SKILL.md files; parse the frontmatter name and description; emit those lines. (Use head -20 per file plus a simple awk/grep for the frontmatter — don't load full bodies.)
- If neither yields anything, emit a single placeholder: <plugin-name> | <plugin-name> | (no skill metadata — see plugin README).
Sort the output for stable diffs.

Report a summary: N plugins pulled (X updated, Y already current, Z skipped), M skill-index rows written.

Safety

Never run git pull outside the marketplace cache directory.

Never modify the inventory file itself — this skill is read-only against inventory.json.

If a plugin directory has uncommitted local changes, skip the pull and surface the warning.

Output style

One status line per plugin during the pull phase, then a totals line, then the index regeneration totals. Don't dump file contents.