The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to create directories and files on the local file system to set up project scaffolding. It incorporates safety mechanisms such as requiring user confirmation before operating in non-empty directories and explicitly stating that it will not overwrite any existing files.- [INDIRECT_PROMPT_INJECTION]: The skill relies on external specification documents (e.g., tech-stack-rule.md, directory-rule.md) to determine the structure and content of the generated project. This ingestion of untrusted data creates a potential attack surface for indirect prompt injection.
Ingestion points: Reads specification files from the docs/standards/ and docs/specs/ directories.
Boundary markers: The instructions include strong directives for the agent to use the documents as the "sole basis" and to "not guess" or provide alternatives not found in the source text.
Capability inventory: Creating directories and writing files such as configuration examples, dependency lists, and entry points.
Sanitization: There is no mention of sanitization or filtering of the content extracted from the specification documents.

do-directory