Use Case: Data Analysis & Reporting

pull metrics · run Python · generate summaries · post on schedule

What you'll build

An agent that pulls data from an API or CSV, runs a Python script to process it, and posts a formatted summary to Slack or Telegram on a schedule — without you touching any code.

Jump to section

Skills you need SOUL.md template Config setup Example: weekly metrics report Scheduled reports Common issues

Skills you need

Skill	What it unlocks	Required?
python-runner	Execute Python scripts inline — pandas, numpy, matplotlib all work	Core
file-reader	Read CSV, JSON, or text files from the local filesystem	Core
http-request	Call REST APIs to pull live metrics — JSON or CSV responses	Core
file-writer	Save analysis output to CSV, JSON, or Markdown	Recommended
chart-generator	Produce PNG charts from data using matplotlib — embeds in reports	Optional
notion-writer	Push summaries to a Notion database as structured pages	Optional

SOUL.md template

SOUL.md — data analysis agent

# Data Analysis Agent

You pull data from files or APIs, analyse it with Python, and produce clear summaries.

## Identity
- Name: DataBot
- Role: Automated data analysis and reporting

## Analysis process
1. Fetch the data (file path or API endpoint)
2. Write a Python script to process it — use pandas for tabular data
3. Execute the script and capture the output
4. Produce a report: key numbers first, then trends, then anomalies
5. Post the report to the configured channel

## Report format
- Lead with the 3 most important numbers in bold
- Use a markdown table for comparisons (max 5 rows, add "Top 5" label if truncated)
- Flag anomalies: anything >2 standard deviations from the past 4-week average
- End with one sentence: what to watch next week

## Python rules
- Use pandas, numpy — do not install new packages
- Handle missing values explicitly (dropna or fillna, state which)
- Print results clearly — the script output becomes the report source
- Save charts to ~/reports/charts/ if generated

Config setup

Python must be installed with pandas and numpy. The python-runner skill uses the system Python.

openclaw.json — data analysis agent

{
  "skills": [
    "official-python-runner",
    "official-file-reader",
    "official-http-request",
    "official-file-writer"
  ],
  "model": "claude-sonnet-4-5",
  "soulPath": "./SOUL.md",
  "channel": {
    "type": "slack",
    "botToken": "xoxb-your-bot-token",
    "appToken": "xapp-your-app-token"
  }
}

Install Python dependencies

pip install pandas numpy matplotlib

Example: weekly metrics report

A complete example — pulling data from a JSON API and producing a weekly summary. Send this prompt to the agent:

"Fetch this week's data from https://api.myapp.com/metrics?range=7d (API key in env: METRICS_API_KEY), analyse daily active users and revenue, flag any days below the 4-week average, and post a summary to Slack"

The agent will write and run a Python script like this internally:

Example Python script the agent generates

import pandas as pd
import numpy as np
import json, os, requests

# Fetch data
resp = requests.get(
    "https://api.myapp.com/metrics",
    params={"range": "7d"},
    headers={"Authorization": f"Bearer {os.environ['METRICS_API_KEY']}"}
)
data = resp.json()
df = pd.DataFrame(data["days"])
df["date"] = pd.to_datetime(df["date"])

# Key metrics
avg_dau = df["dau"].mean()
total_rev = df["revenue"].sum()
peak_day = df.loc[df["dau"].idxmax(), "date"].strftime("%a %b %d")

# Anomalies (below 4-week avg would need more data; flag below this week's avg)
low_days = df[df["dau"] < avg_dau * 0.8]

print(f"**Weekly Summary**")
print(f"Avg DAU: {avg_dau:,.0f} | Total Revenue: ${total_rev:,.2f} | Peak: {peak_day}")
if not low_days.empty:
    for _, row in low_days.iterrows():
        print(f"⚠ Low DAU on {row['date'].strftime('%a')}: {row['dau']:,} (80% threshold)")

Scheduled reports

Add to openclaw.json — weekly report every Monday 9am

{
  "crons": [
    {
      "name": "Weekly metrics report",
      "schedule": "0 9 * * 1",
      "task": "Fetch last 7 days from https://api.myapp.com/metrics?range=7d using METRICS_API_KEY. Analyse DAU and revenue, flag anomalies, post a markdown summary to Slack."
    },
    {
      "name": "Daily CSV check",
      "schedule": "0 8 * * *",
      "task": "Read ~/data/daily-export.csv, check if row count is within 10% of yesterday's file in the same folder. If not, post an alert to Slack."
    }
  ]
}

Common issues

⚠ Python script fails silently

The agent catches script errors and includes them in the response. Ask: "Run the script and show me the full output including any errors." Add explicit print() statements for intermediate values.

⚠ pandas not found

The python-runner skill uses your system Python. Run: pip install pandas numpy matplotlib — or specify the virtualenv path in your SOUL.md: "Python path: /path/to/venv/bin/python".

⚠ API returns paginated data but agent only reads first page

Specify in your prompt: "The API paginates — fetch all pages using the next_cursor field until it is null." The agent will write a loop in the Python script.

⚠ Report posts to wrong Slack channel

Add channel ID to your cron task: "post to #data-reports (channel ID C012AB3CD)". The Slack skill needs the channel ID, not the display name.

Did this guide solve your problem?