The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill facilitates the authoring of imperative orchestration scripts in .claude/workflows/*.js. It provides a specific procedure to validate these scripts by reading the file content, wrapping it in an asynchronous execution context, and performing a syntax check using node --check.
[INDIRECT_PROMPT_INJECTION]: The skill defines a framework where authored workflows can spawn subagents using an agent() hook. This creates a vulnerability surface where data processed by the workflow could contain instructions that influence subagent behavior.
Ingestion points: Data passed as prompts or options to the agent(prompt, opts) function within authored scripts (SKILL.md).
Boundary markers: None identified; there are no instructions for using delimiters or warnings to ignore embedded instructions in the subagent prompts.
Capability inventory: The skill and its associated workflows have access to tools including Write, Read, Edit, and Bash (SKILL.md frontmatter).
Sanitization: The skill does not describe or enforce sanitization of the input passed to subagents or the structured output returned by them.
[COMMAND_EXECUTION]: The documentation includes a Bash one-liner that uses node -e to perform file system operations (fs.readFileSync, fs.writeFileSync) to transform authored scripts into a format suitable for syntax verification.

workflow-create