The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to download external code via git clone from a third-party GitHub repository (synacktiv/php_filter_chain_generator) and subsequently execute a Python script (php_filter_chain_generator.py) to generate exploit payloads.
[COMMAND_EXECUTION]: The skill utilizes several command-line tools for offensive network operations, including curl, ffuf, wfuzz, dotdotpwn, and grep, often involving complex shell pipelines for vulnerability discovery.
[EXTERNAL_DOWNLOADS]: Instructions include fetching content and scripts from external GitHub repositories that are not recognized as trusted organizations in the analysis framework.
[DATA_EXFILTRATION]: The skill provides detailed methodologies and specific file paths for accessing sensitive information on target systems, including /etc/passwd, .env files, wp-config.php, and cloud service credentials located in ~/.aws/credentials.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied data such as target URLs and IP addresses, which are then interpolated into shell commands, creating a potential injection surface.
Ingestion points: The $TARGET variable is used across multiple shell commands throughout the SKILL.md file.
Boundary markers: The skill includes a 'Mandatory confirmation gate' requiring the agent to ask the user for target details and explicit authorization before execution.
Capability inventory: The skill has access to network tools (curl, ffuf, wfuzz), file system tools (cat, grep), and the ability to execute downloaded scripts (python3, git).
Sanitization: No explicit sanitization or escaping mechanisms are defined for the user-supplied target inputs before they are used in shell commands.

hunt-lfi