The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of new agent instructions based on external documentation and user-provided behavior examples.
Ingestion points: Untrusted data is collected from external URLs, local source files, and user-provided feedback or traces as defined in references/synthesis-path.md and references/iteration-path.md.
Boundary markers: The skill instructions do not explicitly mandate the use of delimiters or 'ignore' instructions when interpolating this untrusted data into the final skill artifacts.
Capability inventory: Across the workflow, the agent is directed to write files to the local system (authoring-path.md) and execute shell commands (registration-validation.md).
Sanitization: Although the skill advises anonymizing data and applying trust rules, it does not detail specific methods for escaping or validating the content before it is processed by the AI.
[COMMAND_EXECUTION]: During the registration and validation phase (references/registration-validation.md), the skill directs the agent to execute shell-based validation scripts, specifically npm run chain and npm run catalog.

skill-writer