The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard system utilities such as docker, curl, and git. All user-provided inputs (like application names and tags) are strictly validated using regular expressions before use, effectively preventing shell injection attacks.
[CREDENTIALS_SAFE]: Secret handling is managed with high security standards. The script passes credentials to docker login via standard input to prevent them from appearing in process logs, and it uses a temporary, restricted-permission curlrc file for registry API calls. It also provides clear instructions on avoiding hardcoded secrets and using proper file permissions (e.g., chmod 600).
[DYNAMIC_EXECUTION]: The configuration loading logic is designed for security. Instead of using source or eval which could execute arbitrary code, the script manually parses KEY=value pairs from environment files. This ensures that a malicious configuration file cannot trigger unauthorized code execution.
[DATA_EXFILTRATION]: No unauthorized data transfer was detected. The skill only communicates with the Harbor registry domain explicitly provided by the user in local configuration files or environment variables.
[INDIRECT_PROMPT_INJECTION]: While the skill accepts parameters that could theoretically be used for injection, the implementation uses strict OCI-compliant character whitelists for app names and tags, neutralizing the risk of indirect injection via user-supplied metadata.

xgent-image-push