The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to use npx to fetch and execute the claude-flow package from the NPM registry. While the registry is a well-known service, executing unpinned packages at runtime introduces a dependency on external code that could be subject to supply chain risks.
[COMMAND_EXECUTION]: The agent is prompted to execute shell commands (e.g., npx claude-flow workflow execute) to manage automation tasks, which gives the skill control over local command-line operations.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow structure where natural language tasks are passed to different agents. This creates a surface for indirect prompt injection if malicious instructions are embedded within the workflow steps.
Ingestion points: Workflow YAML configurations and template identifiers.
Boundary markers: Absent; instructions do not specify delimiters to isolate task data.
Capability inventory: Execution of shell commands and coordination of multiple agent roles.
Sanitization: No filtering or validation of workflow task content is described.

workflow-automation