The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to manage dotfiles by executing shell commands such as chezmoi, git, and a bundled Python utility script. These operations are core to the skill's functionality and are documented with clear safety guidelines to prevent unauthorized or destructive actions.
[EXTERNAL_DOWNLOADS]: The official_docs.py script downloads content from the official chezmoi.io website. This is a well-known service for the tool being managed, and the downloads are limited to documentation pages for lookup purposes.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by retrieving and processing data from an external website (chezmoi.io). If the documentation source were compromised, malicious instructions could be ingested into the agent's context.
Ingestion points: The scripts/official_docs.py script fetches documentation from https://www.chezmoi.io/ at runtime.
Boundary markers: The skill treats the fetched content as authoritative documentation, but does not consistently apply explicit delimiters or warnings to ignore instructions embedded within that data.
Capability inventory: The agent has access to powerful CLI tools (chezmoi, git) that can modify sensitive configuration files in the user's home directory.
Sanitization: The ArticleMarkdownParser in the scraper script mitigates risks by stripping HTML tags known for executing code or embedding content, such as <script>, <style>, and <svg>.

better-chezmoi