The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates a 'Security Model' section that explicitly instructs the agent to treat repository files, specifications, and tool logs as data rather than instructions. This is a proactive defense against prompt injection from untrusted source code or documentation.
[DATA_EXFILTRATION]: No network operations or credential-access patterns were detected. The skill specifically states it does not make network actions.
[REMOTE_CODE_EXECUTION]: The skill does not install packages or execute remote scripts. While implementation plans include commands like pytest and git commit, these are presented as templates for a developer to review and use, not for the agent to execute autonomously.
[SAFE]: No obfuscation, persistence mechanisms, or privilege escalation patterns were found. The skill follows standard software development lifecycle practices.

plan-crafting