The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes source code in the repository to determine where formal verification is most needed. This creates a surface where the agent could process untrusted data (code content or comments) containing instructions meant to influence its behavior during the generation of the new skill.
Ingestion points: The skill is instructed to "Answer from the codebase" and "Choose the target from the code."
Boundary markers: No explicit boundary markers or isolation warnings are defined for the code-reading process.
Capability inventory: The skill can create directories, write SKILL.md files, create symbolic links, and perform git commits.
Sanitization: No explicit sanitization or filtering of codebase content is mentioned before it is processed.
[DYNAMIC_EXECUTION]: The skill is designed to programmatically generate and write new AI agent instruction files (SKILL.md) to the local file system. These files are intended to be loaded and executed as local skills by the agent platform.
Evidence: The instructions direct the agent to "Create the real directory .agents/skills//" and "Write the project-local skill there."

create-formal-verification-skill