The Agent Skills Directory

[SAFE]: The skill follows a robust architectural pattern by offloading mechanical tasks, such as topic clustering and document generation, to bundled helper scripts. This separation of concerns simplifies the orchestration prompt and ensures predictable document layout.
[EXTERNAL_DOWNLOADS]: The skill interacts with the Consensus academic search service (consensus.app) and references the 'docx' Node.js package. The generation script includes logic to help the user install the 'docx' dependency via npm if it is not present in the environment. Both are well-known technology services.
[COMMAND_EXECUTION]: The orchestration pipeline invokes bundled local scripts for specific functions: 'citation_tracker.py' (enforcing rate limits and search discipline), 'topic_grouper.py' (clustering topics), 'discussion_question_validator.py' (quality control), and 'generate_reading_list.js' (document assembly). These scripts are bundled with the skill and do not permit arbitrary command injection.
[PROMPT_INJECTION]: The skill processes untrusted syllabus documents (PDF, DOCX, images), creating a surface for indirect prompt injection. This risk is mitigated by the following design elements:
Ingestion points: Phase 1 extraction of content from user-provided files.
Boundary markers: The workflow is gated by an initial 'Grill-Me' intake and a mandatory 'Group-and-Confirm' checkpoint in Phase 2, requiring explicit user approval of topics before Consensus searches begin.
Capability inventory: Access to file reading/writing, Consensus API, and subprocess execution.
Sanitization: The skill separates content generation from layout by passing structured JSON to a dedicated Node.js generator script, providing a layer of data-handling isolation.

syllabus