depends

Map the dependency graph of the Claude Code ecosystem. For any plugin, show exactly what it contributes (skills, agents, MCPs, hooks, commands) and what would break or be lost if it were removed. Answers "what do I lose if I disable X?"

<Use_When>

• User says "depends", "dependency", "what does X give me", "blast radius", "what if I remove"
• Before removing a plugin via /moltbloat:clean
• When deciding between overlapping plugins </Use_When>

<Do_Not_Use_When>

• User wants a full audit — use /moltbloat:audit
• User wants a quick single-plugin lookup — use /moltbloat:why </Do_Not_Use_When>

1. Determine scope

   Check if the user specified a plugin name. If yes, analyze just that plugin. If no, build the full graph for all enabled plugins.

2. Inventory each plugin's contributions

   For each enabled plugin, scan its cached directory to catalog what it provides:

   PLUGIN_DIR=~/.claude/plugins/cache/<marketplace>/<plugin>/<version>
   

   Skills:

   find "$PLUGIN_DIR/skills" -name "SKILL.md" -type f 2>/dev/null | while read f; do
     dir=$(dirname "$f")
     name=$(basename "$dir")
     echo "skill: $name"
   done
   

   Agents:

   find "$PLUGIN_DIR/agents" -name "*.md" -type f 2>/dev/null | while read f; do
     name=$(basename "$f" .md)
     echo "agent: $name"
   done
   

   MCP servers:

   cat "$PLUGIN_DIR/.mcp.json" 2>/dev/null
   

   Parse the JSON to list MCP server names and their tool counts.

   Hooks:

   cat "$PLUGIN_DIR/hooks/hooks.json" 2>/dev/null
   

   List hook types (PreToolUse, PostToolUse, Stop, etc.) and what they trigger.

   Commands:

   find "$PLUGIN_DIR/commands" -name "*.md" -type f 2>/dev/null | while read f; do
     name=$(basename "$f" .md)
     echo "command: $name"
   done
   

   CLAUDE.md instructions:

   wc -c "$PLUGIN_DIR/CLAUDE.md" 2>/dev/null
   wc -c "$PLUGIN_DIR/AGENTS.md" 2>/dev/null
   

3. Detect cross-plugin dependencies

   Build a grep pattern dynamically from all installed plugin names, then check if any plugin's skills or hooks reference another plugin:

   # Build pattern from all installed plugin names (excluding the current one)
   pattern=$(cat ~/.claude/plugins/installed_plugins.json 2>/dev/null | grep -oE '"[^"]+@[^"]+' | sed 's/^"//' | cut -d'@' -f1 | grep -v "<current_plugin>" | paste -sd'|' -)
   grep -rE "$pattern" "$PLUGIN_DIR/skills/" "$PLUGIN_DIR/hooks/" 2>/dev/null
   

4. Build the dependency table

   For a single plugin query:

   # Moltbloat Dependency Report: <plugin-name>
   
   ## Provides
   
   | Type | Count | Items |
   |------|-------|-------|
   | Skills | <N> | <list top skill names> |
   | Agents | <N> | <list agent names> |
   | MCP servers | <N> | <list server names and tool categories> |
   | MCP tools | ~<N> | <list tool name samples> |
   | Hooks | <N> | <list hook types registered> |
   | Commands | <N> | <list or "—"> |
   | CLAUDE.md | <X> KB (~<Y> tokens) | <description> |
   
   ## Blast Radius if Removed
   
   **You would lose:**
   <list each category with counts and notable items>
   
   **Other plugins affected:**
   <list any plugins that reference this one, or "None">
   
   **Overlaps with other installed plugins:**
   <list detected skill/agent/MCP overlaps with other plugins>
   
   ## Token Cost
   <breakdown by CLAUDE.md + MCP tools + skill listings>
   - **Total: ~<X> tokens (<Y>% of context window)**
   

   For a full graph:

   # Moltbloat Full Dependency Graph
   
   | Plugin | Skills | Agents | MCPs | Tools | Hooks | Tokens |
   |--------|--------|--------|------|-------|-------|--------|
   <one row per installed plugin with actual counts>
   | **TOTAL** | **X** | **X** | **X** | **X** | **X** | **~X** |
   
   ## Cross-Plugin Dependencies
   <list any detected references between plugins, or "None">
   
   ## Overlap Map
   <list detected agent/skill/MCP overlaps across all plugins>
   

5. Done

   This skill is read-only. Suggest /moltbloat:clean to act on findings or /moltbloat:why <plugin> for a quick single-plugin summary.