The Agent Skills Directory

[SAFE]: The skill serves as a project initializer and workflow controller. It uses standard file operations to create plan and todo documents.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by collecting user input for preferences (layout engine, competition type) and writing them into a shared plan.md file.
Ingestion points: User preferences gathered in the AskUserQuestions step are written to plan.md.
Boundary markers: The skill does not use explicit boundary markers or 'ignore' instructions when writing user input to plan.md.
Capability inventory: The skill uses Agent to call downstream skills, Write to create files, and acknowledges Bash usage for compilation in later stages.
Sanitization: There is no explicit sanitization or filtering of user input strings before they are persisted.
Note: While the surface exists, the risk is negligible as the input is intended for static configuration and the instructions constrain the agent to ask specific, limited questions.

1start-mathmodel