gemini-json-parsing

Parse Gemini CLI headless output (JSON and stream-JSON formats). Covers response extraction, stats interpretation, error handling, and tool call analysis. Use when processing Gemini CLI programmatic output.

⭐3stars🔀0forks📁View on GitHub🕐Updated Jan 3, 2026

When & Why to Use This Skill

This Claude skill facilitates the seamless integration of Gemini CLI into programmatic workflows by providing specialized parsing for JSON and stream-JSON outputs. It empowers developers to efficiently extract AI-generated content, track detailed token usage and costs, analyze tool-calling performance, and implement robust error handling for automated pipelines and headless environments.

Use Cases

  • Automated Response Extraction: Programmatically retrieve AI responses using optimized jq patterns for integration into custom scripts or web applications.
  • Cost and Usage Monitoring: Calculate precise token consumption (including cached vs. billable tokens) and API costs by parsing model-specific statistics.
  • Performance Optimization: Analyze tool call success rates and execution latency to identify bottlenecks and improve the reliability of agentic workflows.
  • Resilient Error Handling: Interpret structured error types and codes from the Gemini CLI to build robust automation that handles API failures gracefully.
  • Real-time Event Processing: Monitor live JSONL streams to track tool interactions, session initialization, and message deltas during long-running AI tasks.
namegemini-json-parsing
descriptionParse Gemini CLI headless output (JSON and stream-JSON formats). Covers response extraction, stats interpretation, error handling, and tool call analysis. Use when processing Gemini CLI programmatic output.
allowed-toolsRead, Bash

Gemini JSON Parsing

🚨 MANDATORY: Invoke gemini-cli-docs First

STOP - Before providing ANY response about Gemini JSON output:

  1. INVOKE gemini-cli-docs skill
  2. QUERY for the specific output format topic
  3. BASE all responses EXCLUSIVELY on official documentation loaded

Overview

Skill for parsing Gemini CLI's structured output formats. Essential for integration workflows where Claude needs to process Gemini's responses programmatically.

When to Use This Skill

Keywords: parse gemini output, json output, stream json, gemini stats, token usage, jq parsing, gemini response

Use this skill when:

  • Extracting responses from Gemini JSON output
  • Analyzing token usage and costs
  • Parsing tool call statistics
  • Handling errors from Gemini CLI
  • Building automation pipelines

Output Formats

Standard JSON (--output-format json)

Single JSON object returned after completion:

{
  "response": "The main AI-generated content",
  "stats": {
    "models": {
      "gemini-2.5-pro": {
        "api": {
          "totalRequests": 2,
          "totalErrors": 0,
          "totalLatencyMs": 5053
        },
        "tokens": {
          "prompt": 24939,
          "candidates": 20,
          "total": 25113,
          "cached": 21263,
          "thoughts": 154,
          "tool": 0
        }
      }
    },
    "tools": {
      "totalCalls": 1,
      "totalSuccess": 1,
      "totalFail": 0,
      "totalDurationMs": 1881,
      "totalDecisions": {
        "accept": 0,
        "reject": 0,
        "modify": 0,
        "auto_accept": 1
      },
      "byName": {
        "google_web_search": {
          "count": 1,
          "success": 1,
          "fail": 0,
          "durationMs": 1881
        }
      }
    },
    "files": {
      "totalLinesAdded": 0,
      "totalLinesRemoved": 0
    }
  },
  "error": {
    "type": "ApiError",
    "message": "Error description",
    "code": 500
  }
}

Stream JSON (--output-format stream-json)

Newline-delimited JSON (JSONL) with real-time events:

Event Type Description Fields
init Session start session_id, model, timestamp
message User/assistant messages role, content, timestamp
tool_use Tool call requests tool_name, tool_id, parameters
tool_result Tool execution results tool_id, status, output
error Non-fatal errors type, message
result Final outcome status, stats

Example stream:

{"type":"init","timestamp":"2025-10-10T12:00:00.000Z","session_id":"abc123","model":"gemini-2.5-flash"}
{"type":"message","role":"user","content":"List files","timestamp":"2025-10-10T12:00:01.000Z"}
{"type":"tool_use","tool_name":"Bash","tool_id":"bash-123","parameters":{"command":"ls -la"}}
{"type":"tool_result","tool_id":"bash-123","status":"success","output":"file1.txt\nfile2.txt"}
{"type":"message","role":"assistant","content":"Here are the files...","delta":true}
{"type":"result","status":"success","stats":{"total_tokens":250}}

Common Extraction Patterns

Extract Response Text

# Get main response
gemini "query" --output-format json | jq -r '.response'

# With error handling
result=$(gemini "query" --output-format json)
if echo "$result" | jq -e '.error' > /dev/null 2>&1; then
  echo "Error: $(echo "$result" | jq -r '.error.message')"
else
  echo "$result" | jq -r '.response'
fi

Token Statistics

# Total tokens used
echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.total) | add'

# Cached tokens (cost savings)
echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.cached) | add'

# Billable tokens
total=$(echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.total) | add')
cached=$(echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.cached) | add')
echo "Billable: $((total - cached))"

# Tokens by model
echo "$result" | jq '.stats.models | to_entries[] | "\(.key): \(.value.tokens.total) tokens"'

Tool Call Analysis

# Total tool calls
echo "$result" | jq '.stats.tools.totalCalls'

# List tools used
echo "$result" | jq -r '.stats.tools.byName | keys | join(", ")'

# Tool success rate
total=$(echo "$result" | jq '.stats.tools.totalCalls')
success=$(echo "$result" | jq '.stats.tools.totalSuccess')
echo "Success rate: $((success * 100 / total))%"

# Detailed tool stats
echo "$result" | jq '.stats.tools.byName | to_entries[] | "\(.key): \(.value.count) calls, \(.value.durationMs)ms"'

Model Usage

# List models used
echo "$result" | jq -r '.stats.models | keys | join(", ")'

# Model latency
echo "$result" | jq '.stats.models | to_entries[] | "\(.key): \(.value.api.totalLatencyMs)ms"'

# Request counts
echo "$result" | jq '.stats.models | to_entries[] | "\(.key): \(.value.api.totalRequests) requests"'

Error Handling

# Check for errors
if echo "$result" | jq -e '.error' > /dev/null 2>&1; then
  error_type=$(echo "$result" | jq -r '.error.type // "Unknown"')
  error_msg=$(echo "$result" | jq -r '.error.message // "No message"')
  error_code=$(echo "$result" | jq -r '.error.code // "N/A"')
  echo "Error [$error_type]: $error_msg (code: $error_code)"
  exit 1
fi

File Modifications

# Lines changed
echo "$result" | jq '"Added: \(.stats.files.totalLinesAdded), Removed: \(.stats.files.totalLinesRemoved)"'

Stream Processing

Filter by Event Type

# Get only tool results
gemini --output-format stream-json -p "query" | jq -r 'select(.type == "tool_result")'

# Get only errors
gemini --output-format stream-json -p "query" | jq -r 'select(.type == "error")'

# Get assistant messages
gemini --output-format stream-json -p "query" | jq -r 'select(.type == "message" and .role == "assistant") | .content'

Real-time Monitoring

# Watch tool calls as they happen
gemini --output-format stream-json -p "analyze code" | while read line; do
  type=$(echo "$line" | jq -r '.type')
  case "$type" in
    tool_use)
      tool=$(echo "$line" | jq -r '.tool_name')
      echo "[TOOL] Calling: $tool"
      ;;
    tool_result)
      status=$(echo "$line" | jq -r '.status')
      echo "[RESULT] Status: $status"
      ;;
    error)
      msg=$(echo "$line" | jq -r '.message')
      echo "[ERROR] $msg"
      ;;
  esac
done

Quick Reference

What jq Command
Response text .response
Total tokens .stats.models | to_entries | map(.value.tokens.total) | add
Cached tokens .stats.models | to_entries | map(.value.tokens.cached) | add
Tool calls .stats.tools.totalCalls
Tools used .stats.tools.byName | keys | join(", ")
Models used .stats.models | keys | join(", ")
Error message .error.message // "none"
Error type .error.type // "none"
Lines added .stats.files.totalLinesAdded
Lines removed .stats.files.totalLinesRemoved
Total latency .stats.models | to_entries | map(.value.api.totalLatencyMs) | add

Complete Example

#!/bin/bash
# Analyze code and report stats

result=$(cat src/main.ts | gemini "Review this code for security issues" --output-format json)

# Check for errors
if echo "$result" | jq -e '.error' > /dev/null 2>&1; then
  echo "Error: $(echo "$result" | jq -r '.error.message')"
  exit 1
fi

# Extract response
echo "=== Security Review ==="
echo "$result" | jq -r '.response'

# Report stats
echo ""
echo "=== Stats ==="
total=$(echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.total) | add // 0')
cached=$(echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.cached) | add // 0')
models=$(echo "$result" | jq -r '.stats.models | keys | join(", ") | if . == "" then "none" else . end')
tools=$(echo "$result" | jq '.stats.tools.totalCalls // 0')

echo "Tokens: $total (cached: $cached)"
echo "Models: $models"
echo "Tool calls: $tools"

Test Scenarios

Scenario 1: Extract Response

Query: "How do I extract the response from Gemini JSON output?" Expected Behavior:

  • Skill activates on "parse gemini output" or "json output"
  • Provides jq extraction pattern Success Criteria: User receives .response extraction command

Scenario 2: Token Usage Analysis

Query: "How do I track token usage from Gemini CLI?" Expected Behavior:

  • Skill activates on "token usage" or "gemini stats"
  • Provides stats extraction patterns Success Criteria: User receives token calculation jq commands

Scenario 3: Stream Processing

Query: "How do I process Gemini CLI stream-json output?" Expected Behavior:

  • Skill activates on "stream json"
  • Provides JSONL processing patterns Success Criteria: User receives real-time stream processing example

References

Query gemini-cli-docs for official documentation on:

  • "json output format"
  • "stream-json output"
  • "headless mode"

Version History

  • v1.1.0 (2025-12-01): Added Test Scenarios section
  • v1.0.0 (2025-11-25): Initial release