The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common shell commands such as tree, find, grep, cat, and jq to analyze repository structure and extract metadata from configuration files. It also uses the ln command to create symbolic links between AGENTS.md and agent-specific files like CLAUDE.md to ensure instruction consistency. These operations are appropriate for the skill's stated purpose of repository auditing and configuration.
[PROMPT_INJECTION]: The skill processes existing instruction files, such as .cursorrules, CLAUDE.md, and .windsurfrules, which are considered untrusted external data. This creates an indirect prompt injection surface where malicious instructions within those files could potentially influence the agent's migration logic. However, the skill incorporates a multi-wave discovery and audit workflow that reduces the risk of unintended behavior by ensuring architectural grounding before any file modification occurs.

init-agent-config