The Agent Skills Directory

[PROMPT_INJECTION]: No patterns of prompt injection, jailbreak attempts, or safety filter bypasses were detected. The skill uses instructional language to guide the agent in following a 'feature-first' architecture and 'Iron Law' risk assessment.
[DATA_EXFILTRATION]: No sensitive data exposure or exfiltration patterns found. The skill does not access sensitive local files (like .ssh or .aws) and uses standard placeholders for identifiers. It correctly recommends using 'flutter_secure_storage' for secrets in production apps.
[REMOTE_CODE_EXECUTION]: No suspicious remote code execution or untrusted downloads were found. The skill uses standard Flutter ecosystem tools (pub, build_runner) and well-known MCP servers for development tasks.
[DYNAMIC_CONTEXT_INJECTION]: No usage of the '!command' syntax or silent shell execution at load time was found. All command execution is documented as part of the manual development workflow.
[OBFUSCATION]: The content is clear and uses no encoding techniques (Base64, hex, zero-width characters) to hide instructions or URLs.
[INDIRECT_PROMPT_INJECTION]: While the skill involves reading and processing code (an inherent surface for indirect injection), it includes robust guidelines for safe data handling and does not instruct the agent to execute untrusted input blindly.
[PERSISTENCE]: No attempts to establish persistence through shell profiles, cron jobs, or registry keys were detected.
[PRIVILEGE_ESCALATION]: No commands requiring elevated privileges (sudo, runas) or unsafe permission changes were identified.

flutter-mobile