The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the use of the ralph CLI tool for workflow management tasks. \n
Evidence: The skill instructions in SKILL.md and references/commands.md describe executing commands such as ralph hats validate, ralph hats graph, and ralph run to perform its primary functions. \n
Context: This execution is restricted to the Ralph toolset and is used for its intended purpose of validating and running user-defined workflows.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection by reading and processing user-authored YAML configuration files. \n
Ingestion points: The agent is instructed to read existing hat collection files from the .ralph/hats/ directory to explain their structure or propose refactors (see SKILL.md). \n
Boundary markers: The instructions do not specify explicit delimiters or warnings to ignore instructions embedded within the YAML data fields. \n
Capability inventory: The agent can read/write files and execute the ralph CLI, both of which are required for its operation. \n
Sanitization: There is no mention of sanitizing or escaping natural language content found within the YAML fields like description or instructions before the agent processes them.

ralph-hats