safety-validation
Validate changes against .agentignore before commit.
When & Why to Use This Skill
This Claude skill provides a critical safety validation layer for AI-driven development. It automatically intercepts file modification attempts and cross-references them against a `.agentignore` file, ensuring that the agent never alters restricted or sensitive 'forbidden zones' within a codebase. By enforcing these boundaries, it maintains project integrity and prevents accidental corruption of core infrastructure.
Use Cases
- Protecting sensitive configuration: Automatically block the agent from modifying environment variables, API keys, or security credentials stored in restricted files.
- Enforcing architectural boundaries: Restrict AI modifications to specific feature directories while protecting legacy code or core system architecture from unintended side effects.
- Compliance and Auditing: Generate a transparent audit trail in ACTION_LOG.md for every safety check, helping teams meet security standards for AI-assisted workflows.
| name | safety_validation |
|---|---|
| description | Validate changes against .agentignore before commit. |
| short-description | Safety validation |
Purpose
Ensure forbidden zones are never modified.
Steps
- Compare modified paths against
.agentignore. - Stop if any forbidden path is touched.
- Record validation in
ACTION_LOG.md.