token-breakdownlisted

# Token Breakdown Query Splunk for detailed token usage analytics of the current Claude Code session. ## Prerequisites The Splunk MCP server must be configured in Claude Code. This is managed by nix-darwin: - `splunk` MCP server defined in `programs.claude.mcpServers` - Credentials injected via Doppler (`SPLUNK_MCP_ENDPOINT`, `SPLUNK_MCP_TOKEN`) - Self-signed TLS handled by `NODE_TLS_REJECT_UNAUTHORIZED=0` in the MCP server env Verify: `mcp__splunk__splunk_get_info` should return the Splunk version. ## Invocation ```text /token-breakdown [session-id] ``` - No arguments: analyzes the current active session - With session-id: analyzes a specific session UUID ## Workflow ### Phase 1: Session Identification Determine the current session ID and validate it has a useful name. **Step 1a — Find session ID:** If no session-id argument was provided, find the current session: ```bash # Compute encoded project path (slashes become hyphens) encoded_path=$(echo "$PWD" | sed 's|^/|-|; s|/|-|g') # Find the most recently modified JSONL session file session_file=$(ls -t "$HOME/.claude/projects/${encoded_path}/"*.jsonl 2>/dev/null | head -1) # Validate a session file was found if [ -z "$session_file" ]; then echo "No active session found for current project directory: $PWD" exit 1 fi # Extract session ID from filename session_id=$(basename "$session_file" .jsonl) ``` **Step 1b — Read session metadata:** Read line 1 of the session file to extract the slug and version: ```