The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface in the 'Story Ideas Generator' workflow.\n
Ingestion points: The skill reads docs/course-description.md, mkdocs.yml, and docs/learning-graph/learning-graph.csv to extract subject domains and themes.\n
Boundary markers: The instructions do not define the use of delimiters or 'ignore embedded instructions' warnings when processing this external content.\n
Capability inventory: The skill possesses capabilities for writing to the filesystem (docs/stories/story-ideas.md), executing system commands for image verification (sips), and performing network operations via the Google Gemini API.\n
Sanitization: There is no evidence of sanitization, escaping, or validation of the ingested textbook content before it is interpolated into prompts for generating story ideas.\n- [COMMAND_EXECUTION]: The skill uses subprocess calls to execute local system utilities for image and file management.\n
Evidence: scripts/generate-images.py and scripts/verify-images.py utilize subprocess.check_output() to invoke the macOS sips utility for verifying generated image dimensions and aspect ratios.\n
Evidence: scripts/uncomment-images.sh uses sed -i to programmatically uncomment image references in markdown files, which is a standard part of the skill's asset integration workflow.

story-generator