The Agent Skills Directory

[COMMAND_EXECUTION]: The skill dynamically constructs and executes command lines to launch the 1C:Enterprise platform binary (1cv8.exe).
Evidence: The PowerShell script scripts/db-run.ps1 utilizes Start-Process to execute the platform with parameters like /F, /S, /N, and /P.
Evidence: The Python script scripts/db-run.py utilizes subprocess.Popen to perform the same task on compatible systems.
Mitigation: The scripts implement a redaction mechanism (Protect-Secrets in PowerShell and _redact in Python) to ensure that sensitive credentials like passwords are not printed to the console or logs.
[INDIRECT_PROMPT_INJECTION]: The skill automatically searches for and parses a .v8-project.json file in the project directory hierarchy to resolve database connection details and additional arguments.
Ingestion points: Configuration data is read from .v8-project.json using ConvertFrom-Json (PowerShell) and json.load (Python) in the find_project_database and Get-ProjectExtraArgs / project_extra_args functions.
Boundary markers: No explicit boundary markers or "ignore instructions" warnings are applied to the data read from the configuration file.
Capability inventory: The skill can execute the 1C platform binary with arguments derived from this file, including arbitrary additional arguments via v8args or ibcmdargs fields.
Sanitization: The scripts include an Assert-ExtraArgs / assert_extra_args function that checks additional arguments against a blacklist of "owned" keys (e.g., /F, /S, /N, /P) and batch commands to prevent the configuration file from overriding the skill's core execution logic.

db-run