The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No patterns of instruction override, jailbreak attempts, or system prompt extraction were found.\n- [DATA_EXFILTRATION] (SAFE): No network operations, hardcoded credentials, or access to sensitive file paths (e.g., .ssh, .env) are present.\n- [COMMAND_EXECUTION] (SAFE): No shell commands, subprocess calls, or privilege escalation attempts were detected.\n- [REMOTE_CODE_EXECUTION] (SAFE): No remote scripts are downloaded or executed. The skill does not perform any package installations or dynamic code execution.\n- [INDIRECT_PROMPT_INJECTION] (LOW): The skill ingests user input related to project definitions. While it lacks explicit boundary markers, it has no write, execute, or network capabilities, limiting the impact of any embedded instructions to local reasoning and advice only.

scoping-cutting