If you discover a security vulnerability in this project, please do not open a
public issue. Instead, open a GitHub Issue
with the title prefix [SECURITY] — this keeps the report visible to maintainers
while the issue is being resolved.
Please include:
- Description of the vulnerability
- Steps to reproduce
- Potential impact
We aim to respond within 72 hours and resolve confirmed issues within 7 days.
This project is a Claude Code skill (prompt-based). Vulnerabilities include:
- Prompt injection vectors in skill inputs
- Instructions that could cause unintended data exposure or deletion
- Safety check bypasses (e.g. skipping the secrets scan in Step 1)
- File-write instructions targeting paths outside the project directory
- Issues with Claude itself or the Anthropic API
- GitHub API rate limits or authentication edge cases
- Cosmetic issues (wrong color, typo in output)