The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include high-privilege commands, specifically sudo rm -rf ~/.devcontainer-state, which poses a risk of unauthorized or accidental data deletion.
[DATA_EXFILTRATION]: The configuration requirements include mounting sensitive host directories (~/.ssh and ~/.aws) directly into the development container, creating a significant risk of credential exposure or theft.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration state and blueprints from external GitHub repositories such as git@github.com:loxosceles/devcontainer-state.git and loxosceles/project-blueprints.
[REMOTE_CODE_EXECUTION]: The skill uses npx to dynamically download and execute skills and tools (e.g., npx skills add loxosceles/ai-dev) during the migration process.
[PROMPT_INJECTION]: The skill processes existing, potentially untrusted project files, establishing an indirect prompt injection surface.
Ingestion points: The agent reads and audits existing project files, including .devcontainer/ configs, linting files, and package.json (SKILL.md).
Boundary markers: None identified; instructions do not include delimiters or warnings to ignore malicious content within project files.
Capability inventory: The skill has extensive shell execution capabilities including git, mkdir, npx, and pnpm, as well as the ability to modify project structure (SKILL.md).
Sanitization: No validation or sanitization of existing project content is specified before processing.

project-migration