---

name: jira-crash description: "Fetch crash information from Jira issues, extract crash platform links from comments, download ELF/coredump/log files using playwright browser automation, and organize files for GDB debugging. Supports multiple cores (ap, cp, sensor, audio) and auto-detects latest crash from tool comments." license: Apache-2.0 compatibility: "Requires Jira MCP server, playwright MCP server with proxy for internal network access. Optional: executor MCP for GDB sessions."

jira-crash

Fetch crash information from Jira issues and download debug files from crash analysis platform using playwright browser automation.

When to Use

• Analyze crash issues from Jira
• Download ELF, coredump, and log files from crash analysis platform
• Fetch crash files for multiple cores (ap, cp, sensor, audio)

Prerequisites

Required:

• Jira MCP server - Fetch Jira issue information
• Playwright MCP server - Browser automation for crash platform access

Playwright configuration (with proxy if needed):

Copy"playwright": {
  "command": "npx",
  "args": ["@playwright/mcp@latest", "--proxy-server", "http://127.0.0.1:7890"]
}

Workflow

Step 1: Fetch Jira Issue Basic Information

Use mcp_mi_jira_jira_get_issue to get issue details:

Copyissue_key: "PROJECT-XXXXX"
fields: "summary,description,status,assignee,reporter,issuetype,created,priority,labels"
comment_limit: 20

Extract key information:

• summary: Issue title
• description: Issue description containing crash report link
• status: Current status
• labels: Label information

Step 2: Extract Crash Platform Link

From description:

CopyCrash report link format: http://crash.example.internal/crash_report?id=XXXXXX

From comments (important):

Tools automatically comment new identical crash information to Jira. Check recent comments:

• If multiple crashes are auto-commented by tools, get the crash link from the latest comment
• If no tool comments exist, use crash information from the main issue description

Identifying tool auto-comment characteristics:

• Comment content contains crash_report?id= link
• Comment format is typically structured crash information

Step 3: Access Crash Platform Using Playwright

Copymcp_playwright_browser_navigate: http://crash.example.internal/crash_report?id={crash_id}

Authentication handling:

• First access may redirect to SSO login page
• If intermediate page appears, click "Continue to access"
• Use mcp_playwright_browser_take_screenshot to capture QR code for user scanning with authenticator app

Step 4: Extract Download Links

Identify available core types from page (via tab labels):

• ap - Application Processor (most common)
• cp - Communication Processor
• sensor - Sensor Hub
• audio - Audio DSP
• tee - Trusted Execution Environment

URL patterns:

CopyELF:  https://storage.example.internal/crash-logs/{crash_id}/{core}/firmware_{core}.elf
Dump: https://storage.example.internal/crash-logs/{crash_id}/{core}/*.bin
Log:  https://storage.example.internal/crash-logs/{crash_id}/{core}/full_run.log

Step 5: Download Files to Local Directory

Directory structure:

Copy./crash-{crash_id}/
├── ap/
│   ├── firmware_ap.elf
│   ├── 0-0x40000000.bin
│   └── full_run.log
├── cp/
│   ├── firmware_cp.elf
│   ├── *.bin
│   └── full_run.log
├── sensor/
│   └── ...
└── audio/
    └── ...

Download commands:

Copymkdir -p ./crash-{crash_id}/{core}
curl -L -o ./crash-{crash_id}/{core}/firmware_{core}.elf "{elf_url}"
curl -L -o ./crash-{crash_id}/{core}/{dump_filename} "{dump_url}"
curl -L -o ./crash-{crash_id}/{core}/full_run.log "{log_url}"

Output

After successful execution:

• Downloaded file paths
• Crash basic information (product, branch, crash type)
• Command hints for GDB analysis

If any download fails: STOP and ask user - do not proceed Never use existing local files or files from other directories

Examples

Example 1: Standard Jira Crash Analysis

User request:

CopyAnalyze crash issue PROJECT-12345

Execution flow:

1. Get Jira issue:

Copymcp_mi_jira_jira_get_issue(issue_key="PROJECT-12345", fields="summary,description,status", comment_limit=20)

1. Extract crash_id from description or latest comment: 303533

2. Access crash platform:

Copymcp_playwright_browser_navigate(url="http://crash.example.internal/crash_report?id=303533")

1. Get page snapshot to extract download links:

Copymcp_playwright_browser_snapshot()

1. Download AP core files:

Copymkdir -p ./crash-303533/ap
curl -L -o ./crash-303533/ap/firmware_ap.elf "https://storage.example.internal/crash-logs/303533/ap/firmware_ap.elf"
curl -L -o ./crash-303533/ap/0-0x40000000.bin "https://storage.example.internal/crash-logs/303533/ap/0-0x40000000.bin"
curl -L -o ./crash-303533/ap/full_run.log "https://storage.example.internal/crash-logs/303533/ap/full_run.log"

Example 2: Multi-Core Crash

Scenario: Jira comments contain multiple crashes, need to download files for multiple cores

Execution:

Copy# Download all available cores
for core in ap cp sensor audio; do
  mkdir -p ./crash-{crash_id}/$core
  curl -L -f -o ./crash-{crash_id}/$core/firmware_$core.elf \
    "https://storage.example.internal/crash-logs/{crash_id}/$core/firmware_$core.elf" || true
done

Example 3: Get Latest Crash from Tool Auto-Comments

Scenario: Jira has multiple tool auto-commented crashes

Processing logic:

1. Get recent 20 comments
2. Filter comments containing crash_report?id=
3. Take crash_id from the latest matching comment
4. Download files for that crash

Error Handling

Login Issues

If redirected to SSO login page:

1. Screenshot to display QR code
2. User scans with authenticator app
3. Continue after successful login

File Not Found

Some cores may not have crash files, use curl -f to ignore 404 errors:

Copycurl -L -f -o file.elf "url" || echo "File not found, skipping"

Incomplete Coredump

Page may show "coredump incomplete, requires manual analysis":

• Still download available dump files
• Analyze with full_run.log

Integration

After download, use crash-analysis skill or gdb-start skill for debugging:

Copy# Start GDB
gdb-multiarch ./crash-{crash_id}/ap/firmware_ap.elf

# Import nxgdb
(gdb) py import nxgdb

# Load dump
(gdb) target nxstub --rawfile ./crash-{crash_id}/ap/0-0x40000000.bin:0x40000000

URL Patterns

Notes

• Generally prioritize analyzing ap core crash data
• Dump filename format {offset}-{base_address}.bin, e.g., 0-0x40000000.bin means base address 0x40000000
• If multiple crashes are auto-commented to Jira by tools, default to getting the latest one