The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, creating a surface for indirect prompt injection.
Ingestion points: Data enters the system via the defuddle parse command (which fetches content from URLs), the --data flag for local JSON/CSV files, and piped stdin.
Boundary markers: The instructions do not include boundary markers or delimiters to signal that the AI should ignore potential instructions embedded within the processed data.
Capability inventory: The skill has the capability to write files to the local filesystem using the -o and --output-dir flags.
Sanitization: A safe_name filter is mentioned for filename generation, but there is no evidence of sanitization or escaping for the primary data content before it is interpolated into templates.
[EXTERNAL_DOWNLOADS]: The instructions suggest installing the 'knap' CLI tool globally from the NPM registry using npm install -g knap or executing it via npx knap.
[COMMAND_EXECUTION]: The skill functions by directing the agent to run various shell commands for the 'knap' and 'defuddle' utilities to perform rendering, parsing, and validation tasks.
[DYNAMIC_EXECUTION]: The skill uses a templating engine that supports logic (e.g., {% if %} blocks) and variable interpolation to generate Markdown content dynamically at runtime based on the provided input data.

knap