ai-ecom-skills

AI e-commerce skills for Qianchuan (巨量千川) advertising operations.

Compatible with Codex, OpenClaw, and Cursor.

Installation

pip install ai-ecom-skills

Quick Start

# Install skill to all platforms (Codex/OpenClaw/Cursor)
python -m ai_ecom_skills.cli install

# Set backend API URL
export BACKEND_API_BASE_URL=http://localhost:8080  # macOS/Linux
set BACKEND_API_BASE_URL=http://localhost:8080     # Windows

# Restart your agent (Codex/OpenClaw/Cursor)

CLI Commands

python -m ai_ecom_skills.cli install           # Install to all platforms
python -m ai_ecom_skills.cli install codex     # Install to Codex only
python -m ai_ecom_skills.cli install openclaw  # Install to OpenClaw only
python -m ai_ecom_skills.cli uninstall         # Uninstall from all platforms
python -m ai_ecom_skills.cli list              # Show installation status
python -m ai_ecom_skills.cli info              # Show skill information

Features

Analysis Skills

• ReviewSummarySkill - Summarize execution results (before/after comparison)
• RiskGuardrailSkill - Check risk policies (kill_switch, budget limits)
• ActionProposalDraftSkill - Generate optimization proposals

Fetch Skills

• FetchPlanCoreContextSkill - Fetch plan basic info
• FetchPlanMetricsSkill - Fetch daily metrics (ROI/cost/GMV)
• FetchPlanHourlyMetricsSkill - Fetch hourly metrics

Execute Skills

• ExecuteAdjustBudgetSkill - Adjust plan budget
• ExecuteAdjustBidSkill - Adjust plan bid
• ExecutePausePlanSkill - Pause an ad plan

Python Usage

from ai_ecom_skills.skills.review_summary import ReviewSummarySkill
from ai_ecom_skills.schemas.agent import AgentTaskRequest

skill = ReviewSummarySkill()
request = AgentTaskRequest(
    targetId="plan-001",
    context={
        "metrics_before": {"roi": 2.5, "cost": 100.0},
        "metrics_after": {"roi": 3.8, "cost": 120.0},
        "execution_label": "提高预算 20%",
    },
)
suggestion = skill.run(request)
print(suggestion.summary)

Requirements

• Python >= 3.11
• Backend API service running (for Fetch/Execute skills)

License

MIT