The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions that attempt to supersede all other agent behavior: "These rules are non-negotiable and override all other instructions." While used here to enforce documentation quality, this language pattern is a known vector for prompt injection.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to handle and generate content for external documentation files.
Ingestion points: The skill processes README.md, CONTRIBUTING.md, and other documentation files that may incorporate external user input (e.g., in templates.md).
Boundary markers: Absent. There are no instructions for the agent to use delimiters or specifically ignore instructions embedded in the documents it writes or edits.
Capability inventory: The skill describes capabilities including file modification, git repository management, and CLI tool execution.
Sanitization: Absent. The skill does not define methods for sanitizing or validating external data before it is processed or stored in documentation.
[COMMAND_EXECUTION]: Documentation in reference.md suggests using terminal commands such as markdownlint, prettier, and grip. An agent following these instructions might attempt to execute these commands in the shell.
[EXTERNAL_DOWNLOADS]: The skill references several well-known third-party packages for documentation maintenance, specifically markdownlint-cli, prettier, and grip.

markdown-documentation