The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, such as web pages parsed by defuddle or local files, and interpolates this data into Markdown templates.
Ingestion points: Data is ingested via the --data, --data-json, and stdin flags in SKILL.md during the render and batch commands.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore instructions embedded within the processed data.
Capability inventory: The skill has the capability to write to the file system using the -o (output) and --output-dir flags.
Sanitization: The skill documentation recommends using the safe_name filter for data-derived filenames, but general content sanitization for template interpolation is not detailed.
[COMMAND_EXECUTION]: The skill relies on the execution of the knap and defuddle CLI tools to perform its primary functions, including template rendering, validation, and batch processing.
[EXTERNAL_DOWNLOADS]: The instructions recommend installing the knap package globally via npm install -g knap or running it via npx knap to provide the necessary environment for the skill.

knap