The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill provides templates for generating Python scripts (figures/gen_fig_<name>.py) and explicitly instructs the agent to run these scripts to generate diagrams and charts. This pattern of script generation followed by execution is used both for local data plotting with matplotlib/seaborn and for remote image generation via the Gemini API.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted data including paper drafts, experiment results, and structured files (CSV/JSON). This data is used to extract system components and relationships that are directly interpolated into the prompts for image generation and the code for data visualization.
Ingestion points: The workflow relies on processing user-supplied research context, raw data tables, and files like results.csv or data.json (as described in the Context Analysis & Extraction section).
Boundary markers: There are no explicit instructions or delimiters defined to isolate the untrusted content or to prevent the agent from following instructions embedded within the processed text or data.
Capability inventory: The skill possesses the ability to write files to the local system, execute Python code, and perform network requests to the Google Gemini API using an environment-stored API key.
Sanitization: The instructions do not specify any validation, escaping, or filtering of the extracted entities or data values before they are used in script generation or API prompts.
[CREDENTIALS_UNSAFE]: While the skill correctly advises against hardcoding keys and recommends using the GEMINI_API_KEY environment variable, it facilitates a workflow where this sensitive credential is used within scripts whose logic is influenced by untrusted external inputs.

academic-plotting