The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The skill relies heavily on claude-flow@alpha, a third-party package installed via npm install -g or executed via npx. This package is not from a trusted organization and its security posture is unverified.
REMOTE_CODE_EXECUTION (MEDIUM): The extensive use of npx claude-flow@alpha within suggested workflow scripts results in the automatic download and execution of remote code from the npm registry whenever the workflow is run.
COMMAND_EXECUTION (MEDIUM): The skill instructs users to copy a template script (workflow-script-template.sh), grant it execution permissions (chmod +x), and run it. These scripts utilize high-privilege capabilities such as multi-agent orchestration and file editing (agent booster edit).
PROMPT_INJECTION (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8).
Ingestion points: Untrusted data is read from TASK.md in the run_phase1 function.
Boundary markers: No boundary markers or 'ignore' instructions are present when interpolating $(cat TASK.md) into the npx command.
Capability inventory: The skill includes agents with 'development' strategies capable of code changes, and tools for GitHub automation and file editing.
Sanitization: There is no evidence of input sanitization before passing the content of TASK.md to the agent swarm.

creating-workflows