The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from conversation history, which can include untrusted content retrieved from external sources or generated based on malicious inputs.
Ingestion points: The skill is triggered to process "recently researched or generated content" from the conversation context.
Boundary markers: There are no explicit instructions to use delimiters or warnings to ignore embedded instructions within the data being exported.
Capability inventory: The skill utilizes the Write tool to save content to the filesystem and the Bash tool to execute viewing commands.
Sanitization: The instructions focus on formatting (CSS, inline styles, kebab-case filenames) but do not mandate the removal or escaping of potentially malicious HTML tags or JavaScript before writing to the output file.
[COMMAND_EXECUTION]: The skill instructions require the agent to use the Bash tool to open the generated files.
The agent is instructed to execute shell commands (e.g., grip for Markdown files or a browser for HTML) on files that contain untrusted content processed from the conversation history.

export