The Agent Skills Directory

[PROMPT_INJECTION]: The skill enables an indirect prompt injection surface by instructing the agent to read and act upon terminal output from other panes.
Ingestion points: The agent is directed to use herdr pane read and herdr wait output to ingest terminal scrollback and live output into its context.
Boundary markers: The instructions lack delimiters or safety warnings to help the agent distinguish between its own system instructions and potentially untrusted data in the terminal output.
Capability inventory: The agent has high-privilege capabilities including arbitrary command execution via herdr pane run, file system navigation through workspace create --cwd, and environment management.
Sanitization: No sanitization or validation logic is provided to process external terminal data before it is interpreted by the agent.
[COMMAND_EXECUTION]: The skill relies on the herdr CLI tool to perform actions. It also includes a recipe using python3 -c to dynamically parse JSON output from the CLI to extract pane identifiers.
[EXTERNAL_DOWNLOADS]: The herdr integration install command allows the agent to download and install additional modules (e.g., for pi, claude, codex), which involves fetching code or configurations from the tool's backend.

herdr