The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The PowerShell script scripts/epf-dump.ps1 accepts a -Password parameter for 1C:Enterprise authentication and explicitly logs the full command-line arguments, including the password in plain text, to the console using Write-Host "Running: 1cv8.exe $($arguments -join ' ')". This behavior results in sensitive credentials being recorded in the agent's output history and potentially in persistent log files.
[COMMAND_EXECUTION]: The skill determines the path to the 1C:Enterprise executable (v8path) by reading from a .v8-project.json file in the project root. Because this path is used directly in Start-Process without validation or sandboxing, an attacker could supply a malicious path to an arbitrary executable, leading to code execution when the skill is run.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface by ingesting values from an external, untrusted configuration file to influence its execution logic.
Ingestion points: The .v8-project.json file located in the root directory.
Boundary markers: There are no boundary markers, delimiters, or instructions provided to the agent to ignore embedded commands or instructions within the JSON data.
Capability inventory: The skill has capabilities to create and delete directories (New-Item, Remove-Item) and execute external binaries via powershell.exe and Start-Process (scripts/epf-dump.ps1).
Sanitization: There is no evidence of validation, escaping, or filtering of the values retrieved from the JSON configuration before they are interpolated into the PowerShell execution context.

epf-dump