The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains purely instructional content related to Flutter image optimization. It does not include any attempts to override system prompts, bypass safety filters, or use jailbreak-style 'DAN' patterns.
[DATA_EXFILTRATION]: No evidence of data exfiltration was found. The skill includes defensive recommendations regarding private media, advising against placing credentials in URLs and ensuring isolation of user data in caches.
[REMOTE_CODE_EXECUTION]: There are no patterns of downloading and executing scripts from the internet. All references to external resources are links to official documentation (Flutter, Android, Apple, MDN) or well-known ecosystem packages on pub.dev.
[OBFUSCATION]: The content is clear and uses standard technical terminology. No hidden text, zero-width characters, or encoded commands were detected.
[COMMAND_EXECUTION]: The skill does not instruct the agent to execute dangerous shell commands. It focuses on repository inspection and code optimization within the Flutter framework.
[CREDENTIALS_UNSAFE]: No hardcoded API keys or secrets are present. The skill explicitly warns against insecure credential management (e.g., secrets in query strings).
[PERSISTENCE]: The skill does not attempt to modify startup scripts, crontabs, or system services for persistence.
[INDIRECT_PROMPT_INJECTION]: While the skill involves inspecting external repository code (an attack surface for indirect injection), the instructions are structured to perform specific technical audits and do not contain logic that would be easily exploited to execute arbitrary instructions from processed data.
[DYNAMIC_EXECUTION]: No evidence of dynamic code generation or unsafe deserialization (e.g., pickle, eval) was found.
[PRIVILEGE_ESCALATION]: The skill does not request or use elevated privileges (sudo) or attempt to bypass security controls.

flutter-image-performance