The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts such as scripts/pack_verify.py and scripts/edit_and_verify.py that execute the k2f CLI using subprocess.run(). This is the primary intended method for the skill to interact with the K2F layout engine and is considered normal functionality for this type of tool.
[EXTERNAL_DOWNLOADS]: Instructions in SKILL.md and various references guide the user to install the core K2F tools using standard package managers: pip install k2f and npm i @openk2f/k2f. These are well-known registries and the tools are required for the skill to operate.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data to produce rendered documents, creating a potential surface for indirect prompt injection.
Ingestion points: Data enters the system through Markdown files processed by the markdown command and existing .K2F packages processed by k2f unpack.
Boundary markers: While K2F uses a structured JSON schema for document representation, there are no explicit boundary delimiters or "ignore instructions" warnings documented for the data ingestion phase in the provided scripts.
Capability inventory: The skill has capabilities including shell command execution (scripts/pack_verify.py), file system modification (scaffolding and packing), and network operations (scripts/publish.py performs a POST request to a user-defined origin).
Sanitization: The skill relies on the underlying K2F engine and SDK to sanitize content during the compilation and rendering process.

k2f