The Agent Skills Directory

[COMMAND_EXECUTION]: Provides detailed shell instructions for environment configuration, repository management, and tool execution.
[EXTERNAL_DOWNLOADS]: Fetches core components from the official repository and suggests multiple functional extensions from various third-party GitHub accounts (including wondelai, tlehman, 42-evey, Lethe044, Romanescu11, and others) to be installed into the agent's execution path.
[PRIVILEGE_ESCALATION]: Includes instructions for utilizing administrative privileges via sudo to install and manage systemd services on the host machine.
[PERSISTENCE]: Configures persistent agent activity through the creation of system-level services (systemd) and the implementation of scheduled jobs using cron syntax for automated tasks.
[INDIRECT_PROMPT_INJECTION]: The skill defines a significant attack surface by instructing the agent to process data from external messaging platforms (Slack, Telegram), cloud services (Nextcloud), and local tools (Obsidian, GitHub) while maintaining capabilities for command execution and network requests.
Ingestion points: External APIs, messaging platforms, and local file systems (Obsidian vaults).
Boundary markers: No specific delimiters or safety instructions for handling untrusted data processing are included.
Capability inventory: Shell execution, file system access, and network communication via the agent's core and plugins.
Sanitization: No documentation on escaping or filtering external input before processing.
[DYNAMIC_EXECUTION]: References features for autonomous skill generation and execution from literate programming or observed workflows, which involve runtime code generation and interpretation.

awesome-hermes-agent-ecosystem