The Agent Skills Directory

[SAFE]: The skill consists exclusively of Markdown documentation and instructional text. No executable scripts, such as Python or Shell code, are present in the skill files, precluding technical risks like code execution or persistence.\n- [EXTERNAL_DOWNLOADS]: The documentation contains numerous URLs pointing to well-known academic and technology information services.\n
Evidence: Links to reputable sources like arxiv.org, nature.com, pmc.ncbi.nlm.nih.gov, and technologyreview.com are provided as research citations for the writing principles.\n- [PROMPT_INJECTION]: The skill provides instructions to override the model's default writing style and includes an indirect prompt injection surface through text processing.\n
Style override: SKILL.md defines constraints like a 'Vocabulary Kill-List' and 'Never Do' rules to prevent the model from using common AI signatures. These are benign stylistic instructions aligned with the skill's purpose.\n
Ingestion points: User-provided text enters via the $ARGUMENTS variable in SKILL.md and through 'writing samples' in references/prompting-and-workflow.md.\n
Boundary markers: No explicit boundary markers or warnings are used for the input text.\n
Capability inventory: The skill has no capabilities for subprocess execution, file writes, or network requests.\n
Sanitization: No input sanitization is performed.

human-writing