The Agent Skills Directory

[DATA_EXPOSURE]: The skill uses placeholder credentials in an example configuration file (.env.example). These examples, such as 'DATABASE_URL=postgresql://user:password@localhost:5432/myapp', are intended for configuration guidance and do not expose actual secrets.
[REMOTE_CODE_EXECUTION]: Instructions suggest installing standard, well-known libraries (bcrypt, jsonwebtoken, etc.) and using common system utilities (openssl) for secret generation. No malicious script execution or unauthorized remote downloads were found.
[METADATA_POISONING]: The skill includes an analysis report (skill-report.json) as metadata. While this file contains references to future dates and hypothetical model versions, it does not include malicious instructions or attempts to subvert the agent's safety policies.
[INDIRECT_PROMPT_INJECTION]: The skill provides structured instructions for processing user inputs in the resulting authentication system. It emphasizes security controls like input validation and secure token handling, reducing the risk of injection vulnerabilities in implementations based on these templates.

authentication-setup