The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute Python scripts that are generated at runtime. In references/word-export.md, explicit instructions are provided to the user/agent to run the command python gen_<slug>_docx.py after creating the script file.
[DYNAMIC_EXECUTION]: The skill uses Python code templates located in references/word-export.md to dynamically generate executable .py files. These scripts are then executed to generate branded Word reports, representing a runtime code generation and execution flow.
[DATA_EXFILTRATION]: Several files in the skill (notably references/cerv-2026-worked-example.md and references/word-export.md) contain hardcoded absolute Windows file paths. These paths expose the local directory structure and the specific username (AGasser) associated with the environment (e.g., C:\Users\AGasser\OneDrive\1. Ane's PROJECTS\...).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources (grant proposals and donor call documents) and perform operations based on that content.
Ingestion points: Untrusted PDF and text documents provided by the user for scoring (e.g., CERV OG Final Proposal 2026.pdf).
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings specified to separate untrusted document content from the agent's instructions.
Capability inventory: The skill possesses the ability to write files to the disk (Markdown, Python scripts, and Word documents) and execute shell commands to run the generated Python scripts.
Sanitization: The skill lacks explicit sanitization or validation logic for the content extracted from external documents before it is interpolated into reports or used in script generation.

donor-proposal-scoring