The Agent Skills Directory

[PROMPT_INJECTION]: The skill metadata contains a discrepancy, identifying 'K-Dense Inc.' as the author instead of the provided context 'sickn33'. The skill also includes a self-designated 'risk: critical' tag in its frontmatter.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its core data ingestion features.
Ingestion points: Academic search results and metadata are fetched from external platforms like Google Scholar and PubMed via scripts/search_google_scholar.py and scripts/search_pubmed.py.
Boundary markers: The instructions lack explicit delimiters or warnings for the agent to ignore potentially malicious instructions embedded in the academic content it processes.
Capability inventory: The skill allows the agent to execute shell commands and write to the local filesystem using Python scripts (scripts/generate_schematic.py) with input derived from these external sources.
Sanitization: There are no documented sanitization or filtering steps to ensure external data is safe before it influences agent behavior or command arguments.
[COMMAND_EXECUTION]: The skill relies on executing local Python scripts to perform search, extraction, and formatting tasks. The instructions specifically direct the agent to 'automatically' generate and refine schematics using a specific persona ('Nano Banana Pro'), which may lead to reduced user oversight during command execution.
[EXTERNAL_DOWNLOADS]: The skill integrates with well-known and trusted academic services, including the CrossRef API, PubMed E-utilities, and the arXiv API. These network operations are standard for the skill's intended purpose.

citation-management