The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous templates and instructions for executing shell commands locally and remotely via SSH, pueue, and systemd-run. This includes complex piped operations and batch command execution.
[PERSISTENCE]: Recommends the use of systemctl --user enable --now pueued to ensure the job management daemon (pueued) persists on the host across reboots.
[INDIRECT_PROMPT_INJECTION]: The skill documents patterns for interpolating user-controlled parameters such as symbols, thresholds, and date ranges directly into filenames and shell commands.
Ingestion points: Parameters used in job labels and filenames in SKILL.md (e.g., {symbol}_{threshold}).
Boundary markers: Absent; the templates do not use specific delimiters or warnings to prevent the interpretation of instructions embedded in these parameters.
Capability inventory: Extensive local and remote command execution via pueue, bash, ssh, and systemd-run as specified in allowed-tools.
Sanitization: Absent; the skill assumes valid inputs and does not provide patterns for escaping or validating the content of interpolated parameters.
[REMOTE_CODE_EXECUTION]: Deployment protocols include fetching code from remote Git repositories and installing packages via uv pip install from external registries.
[DYNAMIC_EXECUTION]: Suggests the use of eval "$(mise env)" for loading environment configurations and xargs -I{} bash -c '{}' for executing batches of commands stored in a file.

distributed-job-safety