The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions suggest using domain language from a file named GLOSSARY.md. This represents a data ingestion surface where external content is incorporated into the agent's context. However, the skill lacks any active capabilities (such as shell execution or file writing) that could be exploited via this surface.
[EXTERNAL_DOWNLOADS]: The UPSTREAM.md file contains documentation for manual installation and synchronization using npx and shell scripts. These are standard package management and provenance records intended for the user, not instructions for the agent to execute code at runtime.
[SAFE]: The core functionality is restricted to providing markdown templates and formatting examples (Mermaid diagrams, diffs, and trees) to assist in technical documentation.

pr