The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/process.py contains a check_machine_account_quota function that generates a Python script string using f-strings and executes it via subprocess.run using the python3 -c flag. This implementation is vulnerable to code injection if variables such as username or password contain malicious Python code designed to escape the string delimiters.
[COMMAND_EXECUTION]: Both scripts/agent.py and scripts/process.py rely on the subprocess module to call external binaries like noPac.py, powershell.exe, and wmic.exe. These calls utilize parameters (e.g., domain names, IP addresses) that are passed into the skill's execution context without rigorous validation.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and act upon data (credentials, domain names, target IPs) that may be retrieved from untrusted sources. The mandatory evidence chain for this finding is as follows:
Ingestion points: Data enters via command-line arguments in scripts/agent.py and scripts/process.py.
Boundary markers: No delimiters or safety instructions are used to separate untrusted input from the command logic.
Capability inventory: The skill has the capability to execute shell commands and dynamic Python code via the subprocess module.
Sanitization: The skill lacks any sanitization or escaping mechanisms for the interpolated strings, allowing malicious inputs to manipulate the execution flow.
[METADATA_POISONING]: There is an inconsistency in the author metadata; SKILL.md lists 'mahipal' as the author, while the LICENSE file is attributed to 'mukul975'.

exploiting-nopac-cve-2021-42278-42287