The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which introduces a surface for indirect instructions.\n
Ingestion points: SKILL.md outlines procedures for reading reviewer annotations from reMarkable and student survey answers for qualitative analysis.\n
Boundary markers: There are no explicit markers or instructions provided to delimit these external inputs from the agent's primary instructions.\n
Capability inventory: The skill has access to git, gh CLI, canvaslms CLI, and the llm package to manage repository state and external platform data.\n
Sanitization: The skill lacks explicit sanitization or escaping instructions for the external content it processes.\n- [DYNAMIC_EXECUTION]: The skill generates and executes code as part of a literate programming workflow.\n
Evidence: SKILL.md describes the "tangling" of noweb source files into runnable artifacts like Python scripts and platform-specific JSON files, which are then used to create instruments and analyze data.\n- [COMMAND_EXECUTION]: The skill uses various command-line utilities to perform workflow tasks.\n
Evidence: The instructions include the use of git push for repository updates, gh pr create and gh issue create for task management, and canvaslms CLI for platform interaction.

scientific-writing