The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The README.md file contains a command curl -fsSL https://herdr.dev/install.sh | sh which pipes a remote script from an unverified external domain to the shell. This allows arbitrary code execution from a non-standard source.
[EXTERNAL_DOWNLOADS]: The skill references and encourages the download of resources from herdr.dev, which is not a verified vendor or well-known service domain according to the established safety parameters.
[COMMAND_EXECUTION]: The skill uses the herdr CLI extensively to perform system-level operations such as process management, pane creation, and shell command execution.
[COMMAND_EXECUTION]: The skill includes a reference script in references/event-monitoring.md that uses a Python HEREDOC pattern to execute dynamic code for monitoring terminal state and lifecycle events via Unix sockets.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its data processing workflow.
Ingestion points: Terminal output is ingested via herdr pane read and herdr agent read in SKILL.md.
Boundary markers: No delimiters or safety instructions are present to prevent the agent from obeying instructions embedded in the terminal output.
Capability inventory: The skill has access to dangerous capabilities, including running shell commands (herdr pane run) and prompting other coding agents (herdr agent prompt).
Sanitization: There is no evidence of filtering or escaping the ingested content before it is used to determine subsequent agent actions.

herdr