The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/fleet.sh uses the eval command to execute the TEST_CMD string defined in the local configuration file .claude/fleet/config. This allows for arbitrary shell command execution during the landing process.
[COMMAND_EXECUTION]: The skill manages a background daemon process using Bash(run_in_background: true) and implements signal trapping and PID file management to coordinate multiple sessions.
[COMMAND_EXECUTION]: The init command performs automated modifications to the repository's .gitignore file and may execute Git commits to track these changes automatically.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface.
Ingestion points: The scripts/signal.sh script reads and parses file content from user-provided paths (intended for test logs) to determine if a lane is "READY".
Boundary markers: None; the content of the log file is processed without delimiters to isolate it from the signaling logic.
Capability inventory: The orchestrator script scripts/fleet.sh can execute shell commands via eval and perform high-impact Git operations like merge, revert, and rebase based on the signals.
Sanitization: Limited to basic keyword filtering (grep) for failure or error markers within the log files.

fleet-ops