The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Deno runtime to execute local analysis scripts (scripts/prose-check.ts and scripts/rhythm.ts). These scripts are invoked with the --allow-read flag, restricting them to reading the provided text files for analysis purposes without performing broader system modifications or network operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted text provided by the user (e.g., fiction chapters).
Ingestion points: Text is read into scripts/prose-check.ts and scripts/rhythm.ts using Deno.readTextFile or passed as a CLI argument via --text.
Boundary markers: The skill does not use specific delimiters to separate user text from instructions, as the scripts perform statistical analysis rather than directly interpolating the text into new LLM prompts.
Capability inventory: The scripts possess read-only file system access and the ability to output results to the console.
Sanitization: The scripts use string splitting and regular expression matching to count word patterns; the input text is not executed or evaluated as code.
[SAFE]: The skill includes functionality to persist its output in local files such as .prose-style-output.md. This behavior is explicitly documented for the user as a way to maintain work across sessions and does not involve modifying sensitive system configuration files.

prose-style