The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process output from external terminal panes and other AI agents, creating a significant surface for indirect prompt injection.
Ingestion points: The agent uses herdr pane read, herdr agent read, and herdr pane wait-output to ingest data from sub-sessions into its own context.
Boundary markers: The instructions do not require the use of delimiters or specific system instructions to ignore commands that might be present in the terminal output or sub-agent responses.
Capability inventory: The agent has the ability to execute shell commands (herdr pane run), start and prompt other agents, and manipulate terminal layouts.
Sanitization: No sanitization or validation of the ingested terminal/agent content is specified before the agent processes it.
[COMMAND_EXECUTION]: The skill provides the agent with the capability to execute arbitrary shell commands in managed terminal panes.
Evidence: SKILL.md provides instruction for executing commands using herdr pane run <pane-id> "<command>".
[DYNAMIC_EXECUTION]: The skill facilitates a multi-agent orchestration pattern where the primary agent sends instructions to and receives output from sub-agents at runtime.
Evidence: SKILL.md details the usage of herdr agent start and herdr agent prompt to control secondary agents and wait for their lifecycle changes.

herdr