name: artemis-debug-secure
description: Database investigation skill for Jira tickets with secure credential handling. Multi-Agent Swarm for 3x faster parallel execution. Auto-learns from investigations, searches similar tickets, integrates with Jira, and detects anomalies.

Artemis Database Debug Skill (SWARM Edition)

Performance

Features

Multi-Agent Swarm Architecture

                         ┌────────────────────────┐
                         │   COORDINATOR AGENT    │
                         │   (Orchestrates all)   │
                         └───────────┬────────────┘
                                     │
          ┌──────────────┬───────────┼───────────┬──────────────┐
          ▼              ▼           ▼           ▼              ▼
    ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐
    │ DB AGENT │  │JIRA AGENT│  │ ANALYSIS │  │ LEARNING │
    │          │  │          │  │  AGENT   │  │  AGENT   │
    │ -Queries │  │ -Fetch   │  │ -Errors  │  │ -Brain   │
    │ -Pool    │  │ -Similar │  │ -RCA     │  │ -Patterns│
    │ -Screen  │  │ -Comment │  │ -Summary │  │ -Learn   │
    └──────────┘  └──────────┘  └──────────┘  └──────────┘
          │              │           │              │
          └──────────────┴───────────┴──────────────┘
                                │
                   ┌────────────┴────────────┐
                   │    SHARED RESOURCES     │
                   │  • Message Bus (async)  │
                   │  • Connection Pool (3)  │
                   │  • Result Store         │
                   └─────────────────────────┘

4-Phase Parallel Workflow

Claude Workflow (AUTO)

Step 0: Auto-Setup (FIRST TIME ONLY)

Check if venv exists, if not run setup:

SKILL_DIR=~/.claude/skills/artemis-debug-secure
if [ ! -d "$SKILL_DIR/.venv" ]; then
  echo "NEED_INSTALL"
  cd "$SKILL_DIR" && ./setup.sh
fi

This automatically:
- Creates Python virtual environment
- Installs playwright, requests, beautifulsoup4
- Installs Chromium browser
- Prompts for Artemis credentials (if not set)

Step 1: Start Daemon (IMPORTANT - DO THIS FIRST!)

Always start daemon before any queries for 3x faster execution:

cd ~/.claude/skills/artemis-debug-secure
.venv/bin/python3 scripts/artemis_daemon.py status || .venv/bin/python3 scripts/artemis_daemon.py start &
sleep 3

Step 2: Check Credentials

SKILL_DIR=~/.claude/skills/artemis-debug-secure
USER=$(whoami)
if [ -f "$SKILL_DIR/users/$USER/.credentials" ]; then
  echo "READY"
else
  echo "NEED_SETUP"
fi

Step 3: If NEED_SETUP -> Ask User

Use AskUserQuestion tool:

Questions:
1. "Artemis username?"
2. "Artemis password?"

Then save credentials:

SKILL_DIR=~/.claude/skills/artemis-debug-secure
USER=$(whoami)
mkdir -p "$SKILL_DIR/users/$USER"
cat > "$SKILL_DIR/users/$USER/.credentials" << 'EOF'
ARTEMIS_URL=https://artemis.568winex.com
ARTEMIS_USER={username_from_user}
ARTEMIS_PASS={password_from_user}
EOF
chmod 600 "$SKILL_DIR/users/$USER/.credentials"

Step 4: Fetch Jira Ticket (if ticket key provided)

Use mcp__jira__jira_get_issue(ticketKey)
Parse description to extract: username, webId, date, type

Step 5: Search Similar Tickets

Use mcp__jira__jira_search_issues with JQL:
  project = TCP AND summary ~ "{ticket_type}" AND status = Done ORDER BY created DESC
Review past solutions before investigating

Step 6: Run Investigation

cd ~/.claude/skills/artemis-debug-secure
.venv/bin/python3 scripts/investigate.py -t {type} -w {webId} -u "{username}" -k {ticket_key} --learn

Step 7: Add Jira Comment

Use mcp__jira__jira_add_comment(ticketKey, body)
Include: findings, root cause, conclusion, screenshots

Step 8: Output To CAS (500 chars max)

To CAS:
{Root cause}. {Conclusion}. {Details}.

Running Ad-Hoc Queries (IMPORTANT)

Always use daemon or run_query.py for ad-hoc queries - NOT direct Artemis class!

Option 1: run_query.py (Recommended)

cd ~/.claude/skills/artemis-debug-secure
.venv/bin/python3 scripts/run_query.py "SELECT TOP 10 [CustomerId], [Username] FROM [Customer] WITH(NOLOCK)" Main

Option 2: Python with Daemon

from scripts.artemis_daemon import query
result = query("SELECT TOP 10 [CustomerId], [Username] FROM [Customer] WITH(NOLOCK)", "Main")
print(result)

NEVER do this (slow - opens new browser):

# BAD - Don't use direct Artemis class for ad-hoc queries
from scripts.core import Artemis
artemis = Artemis(config)
artemis.connect()  # Opens new browser every time!

Auto-Detect Type

Auto-Detect WebId

See docs/webids.md for full list.

SQL Rules (MUST FOLLOW)

Error Detection (AUTO)

The skill automatically detects these issues:

Detected issues are flagged as ALERTS (critical) or WARNINGS (potential).

Auto-Learn System

After each investigation, the skill learns:
- Pattern: ticket type + root cause + solution
- Indicators: key fields from results (RejectSetting, Status, etc.)
- Frequency: how often each cause appears

Use learned data:

# Show common causes for a type
python3 scripts/investigate.py --show-common promotion

# Output: [5x] SameFP: Fingerprint conflict...
#         [3x] TurnoverNotMet: Wagering requirement...

File Reference

Script Reference

Swarm (Recommended - 3x faster)

Classic (Legacy)

CLI Options

Swarm (Recommended)

python3 scripts/swarm_investigate.py [options]

Required (one of):
  --ticket          Jira ticket key (e.g., TCP-12345)
  OR
  -t, --type        Type: promotion, payment, vip, betting, login
  -w, --webid       WebId number
  -u, --username    Player username

Optional:
  --customer-id     CustomerId (auto-fetched if not provided)
  --headless        Run browsers in headless mode
  --pool-size       Connection pool size (default: 3)
  --benchmark       Run performance benchmark (3 iterations)
  --json            Output results as JSON

Classic (Legacy)

python3 scripts/investigate.py [options]

Required:
  -t, --type        Type: promotion, payment, vip, betting, login
  -w, --webid       WebId number
  -u, --username    Player username

Optional:
  -k, --ticket      Jira ticket key (e.g., TCP-12345)
  -c, --customerid  CustomerId (auto-fetched if not provided)
  --headless        Run without browser window
  --learn           Save learning from this investigation
  --show-common     Show common causes for a type
  -o, --output      Save results to JSON file

Multi-User Support

The users/ folder is gitignored. Each laptop creates its own:
- users/{system_username}/.credentials

When skill is shared, new user's credentials are auto-created on first run.

Output Format

To CAS (500 chars max)

{Root cause}. {Conclusion}. {Details}.

Jira Comment

*DATABASE INVESTIGATION RESULTS*
*Ticket:* TCP-12345
*Player:* username
*WebId:* 20154

*Queries Executed:*
- Customer Info
- Rejection Records
- Balance History

*Key Findings:*
- Rejected: Same Fingerprint
- FP used by: other_player

*ROOT CAUSE:* Fingerprint conflict
*CONCLUSION:* Player's device was used by another account

Example Usage

With Jira Ticket

User: Investigate TCP-92018

Claude:
1. Fetch TCP-92018 via mcp__jira__jira_get_issue
2. Parse: username=player123, webId=20154 (Saffaluck), type=promotion
3. Search similar: mcp__jira__jira_search_issues(JQL)
4. Found 3 similar tickets - TCP-91234 had same root cause
5. Check credentials
6. Run: python3 investigate.py -t promotion -w 20154 -u player123 -k TCP-92018 --learn
7. Auto-detect: Fingerprint conflict
8. Comment results to Jira
9. Output To CAS response

Manual Mode

User: Check promotion for player123 on Saffaluck

Claude:
1. Infer: webId=20154, type=promotion
2. Check credentials
3. Run: python3 investigate.py -t promotion -w 20154 -u player123
4. Output To CAS response