The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like git mv to relocate files while preserving repository history. It also provides a grep command targeted at $HOME/.claude/plugins to help the agent locate the plugin's installation directory by searching for its manifest file.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and restructure untrusted project-level data, creating an inherent attack surface for indirect prompt injection.
Ingestion points: The skill reads and evaluates user-controlled project files including AGENTS.md, .agents/MEMORY.md, .sdd-devkit/settings.json, and various specification documents within the repository.
Boundary markers: The instructions explicitly state that the skill must not re-edit, summarize, or translate user wording, and it must obtain approval for a complete migration plan before modifying any files.
Capability inventory: The skill has the capability to perform file moves using git mv and execute shell-based discovery commands.
Sanitization: The skill implements a manual checkpoint by requiring the agent to present all findings and proposed actions to the user in a single confirmation step.

plugin-migrate