The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill uses a placeholder to ingest the tutorial topic from user input, creating a surface for potential injection.
Ingestion points: The $ARGUMENTS variable in SKILL.md interpolates user-provided text directly into the research instructions.
Boundary markers: The instructions explicitly require the agent to confirm the topic with the user if it is missing and forbid the invention of topics, providing a level of control over the input flow.
Capability inventory: The skill has access to file system read tools (Read, Glob, Grep) and restricted shell commands (Bash limited to sbt and gh-query).
Sanitization: No explicit sanitization is applied to the input string, but the skill's output is structured as a non-executable markdown tutorial.
[EXTERNAL_DOWNLOADS]: The skill references external source code repositories for cloning examples.
Evidence: Template sections in SKILL.md and CHECKLIST.md mention git clone https://github.com/zio/<repo-name>.git.
Context: The referenced organization (zio) corresponds to the skill author's official namespace, representing standard resource acquisition for the library's ecosystem.
[COMMAND_EXECUTION]: The skill instructs the agent to use the sbt build tool to verify the documentation and run examples.
Evidence: Commands such as sbt "docs/mdoc --in docs/guides/<tutorial-name>.md" and sbt "schema-examples/runMain ..." are used for technical verification.
Context: These operations are confined to the project's documentation and example directories and are necessary for the skill's stated goal of ensuring technical accuracy in tutorials.

docs-tutorial