The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches content from external URLs, including research papers, course websites, and GitHub repositories, and processes this data through the ed-lesson-author agent. There are no explicit instructions for the agent to treat this external content as untrusted or to use specific boundary markers to prevent malicious instructions embedded in the source material from affecting the agent's behavior.
Ingestion points: references/workflows/explore.md, references/workflows/create-project.md, and agents/ed-lesson-author.md via WebFetch operations.
Boundary markers: Not present in the workflow instructions when interpolating external content into prompts.
Capability inventory: The skill has access to Write, Bash, WebFetch, and WebSearch tools.
Sanitization: No explicit sanitization or filtering of fetched content is defined before processing.
[DYNAMIC_EXECUTION]: The skill generates TypeScript animation scripts for Manim scenes based on its research of external source material. These scripts are then executed using the manim-web library. While this is the intended primary purpose of the skill, the generation of executable code from untrusted external data sources constitutes a dynamic execution risk.
[COMMAND_EXECUTION]: The skill performs shell command execution using the Bash tool and standard command-line utilities. Specifically, the doctor.md workflow executes npm install at the project root, and the autodidact.md workflow executes npx diffbook. These operations are used for environment setup and project maintenance.

ed