The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script herdr_names.py executes the herdr and git CLI tools using subprocess.run. These calls use argument lists rather than shell strings, which prevents shell injection vulnerabilities.
[DATA_EXFILTRATION]: The skill reads the recent-unwrapped output from terminal panes. This data is shared with the AI agent to facilitate naming, which could lead to sensitive information displayed in a terminal being processed by the agent.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from terminal outputs to decide on renaming actions. 1. Ingestion points: The read_tail function in scripts/herdr_names.py reads data from active terminal panes. 2. Boundary markers: Absent. 3. Capability inventory: Renaming tabs, panes, and agents via the herdr command across scripts. 4. Sanitization: None performed on the content used for naming decisions.
[PRIVILEGE_ESCALATION]: Instructions suggest modifying the user's configuration file ~/.config/herdr/config.toml to adjust UI layout, which is a local configuration change performed with user consent.

herdr-names