The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the herdr and git CLI tools to manage system resources. It executes commands to create/remove git worktrees, manage Herdr workspaces, and start background processes. While these are documented functionalities, they involve direct shell interaction.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection during the Collect workflow, as it reads and summarizes output from background processes which could contain adversarial content.
Ingestion points: Data enters the agent context via the herdr agent read command in SKILL.md during the report collection phase.
Boundary markers: There are no explicit delimiters or instructions to treat the worker report as untrusted data.
Capability inventory: The skill possesses the capability to execute shell commands, manage git repositories, and interact with terminal panes via send-keys.
Sanitization: No filtering or sanitization of the worker's output is specified before the content is summarized for the user.

dispatch