The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes instructions to fetch version metadata from the author's repository at https://raw.githubusercontent.com/abpai/skills/main/versions.json. This is a standard vendor-provided update check.
[COMMAND_EXECUTION]: When an update is detected, the agent is directed to ask the user for permission to execute npx skills update human-writer. This gated execution ensures that external package management remains under the user's control.
[PROMPT_INJECTION]: The skill processes arbitrary user-supplied text with filesystem write and edit tools, representing a surface for indirect prompt injection. The current guidelines do not explicitly include delimiters or instructions to ignore commands hidden within the input text.
Ingestion points: User text provided for humanization (SKILL.md).
Boundary markers: Absent.
Capability inventory: Read, Write, Edit, Grep, Glob tools (SKILL.md frontmatter).
Sanitization: Absent.

human-writer