The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the research text processing pipeline. Untrusted input data is ingested in main.py via the get_input_text function and subsequently interpolated into prompt templates in src/extractor.py and src/generator.py using simple string replacement (replace("{{input_text}}", input_text)). There are no boundary markers or instructions to ignore embedded commands within the prompt templates found in prompts/extraction.txt or prompts/literature.txt. An attacker could embed instructions in a research description to manipulate the LLM's output. While the skill's capabilities are limited to calling the Gemini API and writing a Markdown report via src/reporter.py, this injection could lead to deceptive or malicious content being presented in the generated documentation.

reference-finder