The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text through the $ARGUMENTS variable to define the scope and description of new agent skills. This creates an indirect injection surface where malicious input could influence the generated SKILL.md instructions, potentially affecting the agent's behavior when it subsequently loads the new skill.
Ingestion points: The $ARGUMENTS variable in SKILL.md is used to capture the skill name or idea.
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the interpolated user content.
Capability inventory: The skill is capable of writing files to the skills/ directory, executing shell scripts for validation (make check, scripts/validate_skills.py), and invoking subskills.
Sanitization: The workflow does not specify sanitization or validation routines for user-supplied descriptions before they are written to the filesystem.
[COMMAND_EXECUTION]: The workflow instructs the agent to run repository-specific validation tools, such as make check or scripts/validate_skills.py, to ensure the generated skill meets formatting and quality standards.
[EXTERNAL_DOWNLOADS]: The documentation references and suggests using the skill-creator plugin from the Anthropic marketplace to perform automated benchmarking and quality measurements.

add-skill