The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands such as ls and grep to discover installed skills within the user's environment. It also executes bundled Python scripts (validate_workflow_spec.py and validate_skill_md.py) to perform structural and formatting validation on the generated workflow files. These operations are restricted to the local development environment.
[EXTERNAL_DOWNLOADS]: The skill communicates with https://skills.sh/ to retrieve a list of available community skills and their descriptions. While it provides users with npx commands to install these external dependencies, it explicitly forbids automated installation, maintaining a human-in-the-loop requirement for all third-party code additions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external content from a skill registry and user-provided ideas to generate new instructions. To mitigate potential injection risks, it uses a multi-stage process where a 'workflow spec' (Source of Truth) must be generated and confirmed by the user before any actual skill code is produced. Additionally, it enforces structural validation on the ingested metadata to ensure it conforms to expected YAML schemas.

workflow-creator