The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute powershell.exe which runs a local script db-repo.ps1. This chain of execution allows the agent to run complex system commands.
[DYNAMIC_EXECUTION]: The skill's scripts (db-repo.ps1 and db-repo.py) dynamically construct command-line arguments for the 1cv8.exe binary. A parameter AdditionalV8Arguments allows passing arbitrary flags to the executable.
Mitigation: The scripts implement an Assert-ExtraArgs function that blacklists specific command-line keys (e.g., /F, /N, /P, and administrative repository commands) to prevent users from overriding the skill's intended operation or hijacking credentials.
Risk: Despite filtering, dynamic construction of arguments for complex third-party binaries remains a potential attack surface for command injection if the filtering logic is bypassed.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to parse the output of the 1C platform (repo_log.txt) and provide instructions to the agent based on that output.
Ingestion points: The agent reads the text content of the platform's execution log returned by the script.
Boundary markers: The scripts use clearly labeled blocks such as --- Вывод платформы --- and --- End --- to delimit tool output.
Capability inventory: The skill has access to Bash, Glob, Read, and the ability to execute the 1C repository tool.
Sanitization: The output is printed as raw text; if a malicious actor can influence repository metadata (e.g., commit messages or object names), they could attempt to inject instructions that the agent might follow in subsequent steps (e.g., suggesting specific file modifications).
[SAFE_PRACTICE]: The scripts include redaction logic (Protect-Secrets in PS and _redact in Python) to prevent literal password values passed as arguments from being printed in the output logs.

db-repo