The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes makecli commands to perform high-privilege operations such as app deployment, resource deletion, and production environment promotion. These operations are consistent with the tool's primary purpose.
[EXTERNAL_DOWNLOADS]: The installation and update processes fetch the core CLI tool from the vendor's official Homebrew tap and update components through vendor-managed channels.
[REMOTE_CODE_EXECUTION]: The skills management functionality utilizes npx to fetch and execute extensions from the vendor's repository, representing a dynamic code execution path controlled by the vendor.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local YAML/JSON configuration files and OCR processing results, creating a potential surface for indirect prompt injection where malicious content in processed files could influence agent behavior.
Ingestion points: makecli integration ocr (external files), makecli apply (YAML DSL), makecli record create (JSON data), makecli app deploy (Git HEAD).
Boundary markers: The instructions do not define specific prompt delimiters to isolate untrusted data from instructions.
Capability inventory: The skill has capabilities for platform resource management, file system writes, tool installation via npx, and code deployment via Git.
Sanitization: The tool employs schema validation for DSL resources and record data to ensure structural integrity.
[DYNAMIC_CONTEXT_INJECTION]: The instructions utilize the "!" shell prefix syntax to guide the agent in facilitating interactive browser-based authentication (OAuth) and manual configuration steps that require user participation.

makecli