The Agent Skills Directory

[SAFE]: The skill explicitly uses permissionMode: plan when calling the Claude Code tool. This is a security best practice that prevents the tool from making unauthorized file modifications or executing code during the planning phase.
[SAFE]: The skill policy in agents/openai.yaml disables implicit invocation (allow_implicit_invocation: false). This prevents the skill from being triggered automatically by the agent, mitigating the risk of recursive loops or unrequested project mutations.
[SAFE]: Analysis of the prompt templates confirms they are used for legitimate orchestration and reasoning tasks. The instructions provided to the model are transparent and intended to improve plan quality.
[SAFE]: No obfuscated content, hardcoded credentials, suspicious network activity, or unauthorized file access patterns were identified in the skill files.

claude-plan