The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages and executes local commands through operations like tool/run for tool installations and skills/update which delegates to an external Skills CLI tool.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration schemas and usage examples from the author's official GitHub repository (github.com/logbookfordevs/ai-field-kit).
[INDIRECT_PROMPT_INJECTION]: The skill has attack surfaces for processing external data, which are addressed by instructional mitigations.
Ingestion points: Remote JSON URLs via stack/preview, Markdown rule definitions via rules/save, and configuration archives via bundle/import.
Boundary markers: Instructions strictly mandate that the agent must inspect previews, validate schemas, and read back the resulting state to verify changes.
Capability inventory: Includes filesystem writes for rule synchronization (rules/sync), link management for profile activation, and arbitrary command execution for tool management (tool/run).
Sanitization: The instructions explicitly forbid interpolating content into shell commands, requiring serialization as data via JSON files or stdin (--input -) to prevent command injection.

afk-cli