The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the codewhale CLI toolset (status, inspect, logs, artifacts, interrupt, restart, stop) to manage worker states and perform actions like restarting or stopping fleet runs.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by processing data from external worker runs.
Ingestion points: Untrusted data enters the agent context through the outputs of codewhale fleet logs <worker-id> and codewhale fleet artifacts <worker-id> as specified in the triage loop.
Boundary markers: The skill uses instructional boundaries, such as limiting log excerpts to three lines and summarizing artifact references rather than full payloads, to mitigate but not eliminate the risk of the agent obeying instructions embedded in logs.
Capability inventory: The agent has the authority to execute administrative fleet commands (restart, stop, interrupt) based on its evaluation of the ingested data.
Sanitization: There is no mention of technical sanitization or filtering of the log/artifact content; the skill relies on the agent's ability to follow 'Safe Escalation' and 'Authority Boundary' guidelines to prevent secret leakage.

fleet-manager