The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The installation guide in the README describes cloning the project from a GitHub repository and running a local setup script. These resources originate from the author's own infrastructure, which is consistent with the skill's distribution model.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided "rough prompts" to generate optimized versions, creating a theoretical attack surface for indirect injection.
Ingestion points: User input provided via various trigger phrases such as "improve this prompt: [input]" as defined in SKILL.md and README.md.
Boundary markers: The skill lacks explicit delimiters to distinguish untrusted user input from its internal instructions.
Capability inventory: The skill has no access to dangerous tools or capabilities; it only outputs plain text inside Markdown code blocks.
Sanitization: There are no documented sanitization or filtering steps for the processed input.
[COMMAND_EXECUTION]: The operational logic defined in SKILL.md does not utilize any shell commands or system calls. All command-line examples in the documentation are for the user's initial configuration and are not executed by the agent at runtime.

prompt-engineer