The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted text data such as historical literary corpora, digital archives, and OCR results.
Ingestion points: Text data is processed in references/methods-and-examples.md (topic modeling, sentiment analysis) and references/dh-tools-guide.md (OCR processing).
Boundary markers: There are no explicit instructions or delimiters provided to the agent to prevent it from following commands that might be embedded within the analyzed texts.
Capability inventory: The skill utilizes Read, Write, Edit, and Bash tools. It includes script functionality for writing files (pyLDAvis.save_html), reading images (cv2.imread), and executing shell commands (kraken, tesseract).
Sanitization: No specific sanitization or filtering of the ingested text is described before processing.
[EXTERNAL_DOWNLOADS]: The skill references downloading pre-trained models and specialized software from external sources.
Evidence: references/dh-tools-guide.md suggests downloading OCR models from zenodo.org.
Context: Zenodo is a reputable research repository hosted by CERN. Other references target standard package registries (PyPI, CRAN) for libraries like spaCy, NLTK, and stylo.
[COMMAND_EXECUTION]: The skill provides instructions for executing shell-based tools for OCR and stylometry.
Evidence: references/dh-tools-guide.md and references/methods-and-examples.md include command-line examples for tesseract, kraken, and ketos, as well as R scripts for the stylo package.
Context: These tools are standard and necessary components of the Digital Humanities workflows described.

alterlab-digital-humanities