The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on external data (other skill files in the repository) to perform analysis, validation, and upgrade suggestions. This creates a standard indirect prompt injection surface common to developer tools.
Ingestion points: The scripts scripts/analyze-skill.mjs, scripts/upgrade-skill.mjs, and scripts/quick-validate.mjs read and parse SKILL.md files from the repository.
Boundary markers: The instructions do not define explicit boundary markers or instructions to ignore content when the agent is tasked to revise or analyze existing skills.
Capability inventory: The skill includes file-writing capabilities through scripts/init-skill.mjs (creates new skills) and scripts/package-skill.mjs (generates ZIP archives).
Sanitization: There is no explicit sanitization of prompt-like content from the files being analyzed before the agent processes them.

skill-creator