data-agent-rules
github.com/idk-arsh/data-agent-rules- Category
- AI Agents
- Rank
- No. 2243Tools index
Previous survey · No. 2137 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Data Processing
- Interfaces
- Agent Skill / Plugin
- GitHub
- 2 stars
- Date
About
data-agent-rules is a set of rules, skills, and a destructive-SQL guard for AI coding agents including Claude Code, Codex, Cursor, Gemini CLI, and Copilot. It aims to stop agents from scanning, overwriting, or inventing data in a warehouse. The README reports a 162-run DuckDB eval in which Claude Haiku 4.5 without the rules deleted every row of an orders table, mis-scaled prices, and fabricated customer countries.
At a glance
data-agent-rules gives coding agents instructions and supporting tools for safer warehouse work. It is worth trying when you want the agent to preview affected rows, estimate query cost and ask before a destructive change.
Good fit
Data teams using coding agents against development warehouses and able to test their behavior on fixtures.
Consider something else when
Use database permissions when writes must be prevented regardless of what the model decides.
Start here
Ask an agent to remove duplicates from a disposable table. Check whether it identifies affected rows and waits for approval before writing.
Cost and requirements
The MIT-licensed project documents rules, client integrations and a local DuckDB evaluation. Model or warehouse usage may have separate costs.
Limits to know
Instruction and hook coverage varies by client. Keep production credentials read-only or scoped as needed, and check what any preview tool reveals.
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.