The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to determine the Python executable path by reading a 'python' key from a local configuration file (~/.redshift-skill/config.json). This introduces a significant command injection vector, as a modified configuration file could redirect the agent to execute arbitrary binaries or malicious commands instead of the intended Python interpreter.- [CREDENTIALS_UNSAFE]: The skill explicitly requests the agent to read ~/.redshift-skill/config.json. Configuration files for database exploration tools often contain sensitive credentials, including AWS access keys, database passwords, or connection strings, which would be exposed to the agent's environment.- [NO_CODE]: According to the local.remove file, the scripts directory has been removed from the skill distribution. The instructions in SKILL.md still direct the user and the agent to use scripts like setup.py that no longer exist within the package, which is a deceptive practice.

redshift