The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs administrative actions using sudo. It modifies the system /etc/hosts file to add target hostnames (workflow.md) and manages the OpenVPN service, including starting the daemon and terminating processes (vpn-setup.md).- [REMOTE_CODE_EXECUTION]: The skill implements a self-modifying logic loop where the /skill-update command is used to rewrite the skill's own reference documentation and instructions based on "lessons learned" during automated exploitation sessions (skill-improvement.md, workflow.md).- [EXTERNAL_DOWNLOADS]: The skill configuration for browser automation uses npx @playwright/mcp@latest. This command fetches and executes the latest version of the Playwright MCP server from the public npm registry at runtime without version pinning (cloudflare-bypass.md).- [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface as it ingests untrusted data from external platforms that is then processed by agents with significant system capabilities.
Ingestion points: Challenge descriptions, machine metadata, and tech-stack information are read from the HackTheBox platform (platform-navigation.md).
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are used when agents process target data.
Capability inventory: Agents have access to high-privilege shell commands (sudo), system file modification, and the ability to update the skill's own instruction set.
Sanitization: No sanitization or validation is specified for the techniques and lessons extracted from exploitation logs before they are incorporated into the skill's permanent reference files.

hackthebox