The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several system commands to manage its lifecycle. It runs a Python script to scan for existing projects, uses git commands (pull and clone) to fetch updates, and employs cp and rm for file system maintenance during the update process.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to raw.githubusercontent.com to verify version metadata and downloads updated source code from its official repository on github.com. These operations target well-known services and vendor-owned infrastructure.
[DYNAMIC_EXECUTION]: The builder dynamically executes a local Python utility (scan_apps.py) to determine project routing and identify available updates, integrating script output into its decision-making process.
[INDIRECT_PROMPT_INJECTION]: The skill processes project configuration files (hap-plan.json) from the local workspace to restore state or route users. This creates a surface where malformed or malicious data in these local files could potentially influence the agent's behavior. \n
Ingestion points: Configuration files located in the apps/ directory of the PROJECT_ROOT are read by the scan_apps.py script. \n
Boundary markers: Data retrieved from local files is treated as structured JSON but lacks explicit boundary markers or instructions to the agent to ignore embedded commands. \n
Capability inventory: The skill has access to the file system, network, and shell command execution. \n
Sanitization: The skill relies on standard JSON parsing for configuration data without additional sanitization layers to filter out natural language instructions embedded in data fields.

hap-mcp-app-builder