The Agent Skills Directory

[DYNAMIC_EXECUTION]: The PowerShell script (scripts/db-repo.ps1) uses Add-Type to compile an embedded C# class (PlatformProcess) at runtime.\n
Evidence: Source code for the PlatformProcess class is defined as a string and compiled during execution to interface with low-level Windows APIs (kernel32.dll).\n
Purpose: The code is used to manage child process handle inheritance, preventing the agent from hanging when the 1C platform launches background processes.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes output from the 1C platform which could theoretically contain malicious instructions if the repository data is compromised.\n
Ingestion points: Console output and log files from 1cv8.exe are read and parsed in scripts/db-repo.ps1 and scripts/db-repo.py.\n
Boundary markers: Platform output is clearly labeled and separated in the display using specific delimiter headers.\n
Capability inventory: The skill can execute various 1C platform commands, manage repository users, and perform file system operations (scripts/db-repo.py, scripts/db-repo.ps1).\n
Sanitization: The scripts parse the logs for specific status messages but do not perform deep sanitization of the raw output text.\n- [COMMAND_EXECUTION]: The skill executes shell commands and the 1C platform executable using parameters provided by the user.\n
Evidence: Both scripts use subprocess.Popen or Process.Start to run 1cv8.exe based on arguments derived from user input or project configuration.\n
Mitigation: The scripts include logic to redact passwords and secrets from the command-line display to prevent accidental credential exposure.

db-repo