The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze, revise, and review other agent skills, which involves ingesting untrusted instruction sets and scripts.
Ingestion points: Found in SKILL.md and references/package.md, where the agent is instructed to review external SKILL.md files, references, and associated scripts.
Boundary markers: The instructions lack explicit guidance on using delimiters or safety warnings to ignore instructions contained within the skills being analyzed.
Capability inventory: The skill includes validation steps in SKILL.md ("Run changed scripts") and references/package.md ("scripts execute successfully"), which suggests the agent is expected to execute code associated with the reviewed skills.
Sanitization: There are no instructions for sanitizing or escaping content from analyzed skills before processing.
[EXTERNAL_DOWNLOADS]: The skill references policy documentation from well-known technology providers to guide package validation.

writing-great-skills