The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive interaction with the local environment using the herdr CLI and the field.py script. It executes commands to manage terminal panes, create git worktrees, and monitor process states via the Monitor tool and Python's subprocess.run module.
[INDIRECT_PROMPT_INJECTION]: The skill architecture establishes a tight feedback loop where a 'handler' agent is responsible for reading and acting on content generated by 'field' agents.
Ingestion points: The handler agent reads reports from sub-agents using herdr agent read as part of the Triage workflow in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or security warnings to help the handler agent distinguish between data and instructions when reading reports from field agents.
Capability inventory: The handler has the capability to execute shell commands, manage git repositories, and spawn additional agents based on the triaged reports.
Sanitization: No sanitization or validation logic is present to filter malicious instructions that a compromised or adversarial field agent might include in its completion report.

field-agent