The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the herdr CLI tool to manage terminal panes and workspaces. It executes several subcommands: herdr pane current to identify the active session, herdr tab get to retrieve metadata, herdr workspace list to locate the target workspace, and herdr pane move to perform the final action. All operations are confined to the local terminal management environment as specified by the HERDR_ENV=1 requirement.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the outputs of previous herdr commands to construct the final move command. This creates a surface where external labels (tab or workspace names) are processed by the agent.
Ingestion points: The skill reads data from the JSON-like output of herdr pane current, herdr tab get, and herdr workspace list (SKILL.md, steps 1-3).
Boundary markers: The instructions direct the agent to specific fields (e.g., result.pane.pane_id, result.tab.label) which provides structure, though no explicit security boundaries are set to ignore instructions embedded within those labels.
Capability inventory: The skill is capable of executing shell commands to modify the terminal workspace and pane layout.
Sanitization: The skill instructs the agent to wrap the retrieved tab name in double quotes (--label "<preserved_name>") when executing the move command, providing basic shell command protection against whitespace or special characters.

backlog