The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses bun run to execute local TypeScript scripts (scripts/prepare-humanize-input.ts and scripts/verify-change-rate.ts) for text processing and quality control. These executions are scoped to local file paths provided by the user and are intended for the skill's primary function of text analysis and rewriting.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided text for 'humanization' and translation tasks.
Ingestion points: User-supplied strings via the --text argument or file contents via the --file argument in the prepare-humanize-input.ts script.
Boundary markers: The script wraps metadata and cleaned text in specific HTML comments () to separate processing data from content.
Capability inventory: The skill can read and write files within the local environment (specifically in the _workspace directory) and generate static HTML artifacts for explanations.
Sanitization: The sanitizeText function in prepare-humanize-input.ts proactively removes invisible/special characters (zero-width spaces, etc.) and common AI chatbot headers/footers to normalize input and mitigate accidental instruction following.
[OBFUSCATION]: The scripts/prepare-humanize-input.ts script contains defensive regex patterns to detect and remove hidden characters (Unicode tags, zero-width spaces, etc.) from input text, acting as a security measure to ensure input cleanliness before processing.

etc-lgnh