accidental-data-loss-prevention

Blocks irreversible deletion commands until explicit user consent is obtained.

Updated Aug 26, 2026
One-click install
npx skills add https://github.com/Lathika-laa/Recipe_Box --skill accidental-data-loss-prevention-lathika-laa
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: accidental-data-loss-prevention
Source: https://github.com/Lathika-laa/Recipe_Box/tree/main/.github/.gemini/skills/accidental-data-loss-prevention
Command: npx skills add https://github.com/Lathika-laa/Recipe_Box --skill accidental-data-loss-prevention-lathika-laa

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI agents can execute destructive commands like DROP TABLE, TRUNCATE, or gsutil rm without pausing, risking permanent loss of production data or critical infrastructure. This Skill enforces a mandatory stop-and-verify checkpoint before any irreversible deletion runs. ## Core Features & Use Cases - Destructive SQL Guard: Intercepts DROP TABLE/VIEW/SCHEMA/DATABASE, TRUNCATE, and broad DELETE statements missing a WHERE clause or using 1=1. - Cloud Resource Protection: Covers gsutil rm, gcloud storage rm, gcloud projects delete, and deletion of Spanner, BigQuery, Dataproc, secrets, or KMS keys. - Mandatory Consent Workflow: Halts execution, explains the impact and rationale, and waits for explicit affirmative approval before proceeding. - Use Case: Before an agent runs gcloud projects delete my-prod-project, it stops, explains the consequences, and only proceeds after you confirm in the conversation. ## Quick Start Ask the agent to clean up old database tables and it will pause for your explicit approval before running any destructive deletion command.

Frequently Asked Questions about accidental-data-loss-prevention

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent an AI agent from deleting production data?▼

Use a consent-gating skill that halts any irreversible deletion command before execution. The agent must explain the impact, state why deletion is needed, and wait for your explicit approval before proceeding.

What SQL commands require confirmation before execution?▼

DROP TABLE, DROP VIEW, DROP SCHEMA, DROP DATABASE, TRUNCATE, and broad DELETE statements require confirmation. A DELETE is considered broad when it lacks a WHERE clause or uses a tautology like 1=1.

Does this cover Google Cloud resource deletion?▼

Yes, it covers gsutil rm and gcloud storage rm on critical buckets, gcloud projects delete, and deletion of Spanner, BigQuery, and Dataproc resources. It also guards secrets deletion and KMS key destruction.

What happens when a destructive command is detected?▼

Execution halts immediately and the command is not run. The agent explains the deletion impact and its reasoning, then waits for clear affirmative consent in the conversation before proceeding.

When should I not rely on consent prompts alone?▼

Consent prompts are a last line of defense, not a substitute for backups, IAM restrictions, or deletion protection flags. Combine them with infrastructure-level safeguards for production environments.