The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script (run.py) to manage local state files. This script is intended to be invoked by the agent or user via the CLI to track progress, which is a standard pattern for stateful skills. The script operations are limited to reading/writing its own state file in ~/.openclaw and parsing provided step definition files.
[DATA_EXPOSURE]: The skill manages a state file at ~/.openclaw/skill-state/workflow-orchestration/state.yaml. This file contains workflow metadata (step status, instructions, skill names) and does not inherently access or store sensitive credentials or system files.
[INDIRECT_PROMPT_INJECTION]: The skill processes workflow definitions which include instructions for subsequent steps. While this is an ingestion point for instructions, the skill is designed to orchestrate agent actions based on these instructions. The attack surface is limited to the local workflow definitions provided by the user.

workflow-orchestration