agent-migration-plan

This document is a migration plan and agent template specification rather than executable application code. The file itself contains low-risk shell snippets (directory check and find) and no direct network calls, downloads, or hardcoded secrets. Primary risks stem from capability declarations: multiple agents are allowed Bash, Write, Edit, and GitHub CLI usage which, in a real agent runtime, give the agent the ability to execute arbitrary commands, modify files, and perform repo operations. The opaque mcp__claude-flow__* tools referenced are unknown in behavior and could accept or forward credentials or data depending on implementation. Overall the content is not overtly malicious, but it defines a broad attack surface if the agent platform grants these agents unrestricted execution and network privileges. Recommend: treat this skill as requiring moderation — ensure runtime enforces least-privilege (confirm which agents truly need Bash/Write), require explicit per-action user approval for any destructive operations, and audit implementations of mcp__claude-flow__* tools and memory APIs before enabling these agent capabilities.