The Agent Skills Directory

[SAFE]: The skill defines and enforces strict security standards for generated plugins, specifically banning hardcoded credentials and requiring secure network protocols for integrations.- [PROMPT_INJECTION]: Built-in evaluation criteria (evals.json) explicitly verify that the system rejects malicious inputs designed for command or template injection, such as shell metacharacters in naming fields.- [COMMAND_EXECUTION]: Filesystem operations and script executions are limited to local scaffolding and validation tasks. The instructions emphasize quoting bash variables in scripts to prevent word splitting and globbing vulnerabilities.- [DATA_EXFILTRATION]: No evidence of unauthorized data access or external transmission was found. The skill focuses on local project structure and component initialization.

create-plugin