The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains instructions for the agent to execute local Python scripts using the Bash tool to generate visual schematics. The suggested command pattern, python scripts/generate_schematic.py "your diagram description", interpolates user-provided natural language descriptions directly into a shell command. This creates a potential surface for command injection if the user input contains shell metacharacters and is not sanitized by the agent before execution.
[EXTERNAL_DOWNLOADS]: The references/nstc_guidelines.md file encourages the acquisition of external code and templates. It provides instructions to use the tlmgr package manager (a well-known TeX Live service) and to clone a specific third-party GitHub repository (github.com/L-TChen/nstc-proposal) to obtain LaTeX formatting tools. While the package manager is standard, the third-party repository is a non-vendor external source.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user research data to generate structured grant narratives and diagrams.
Ingestion points: User-supplied research descriptions and diagram prompts in SKILL.md.
Boundary markers: None observed; user input is directly interpolated into technical workflows.
Capability inventory: The skill has access to Bash, Write, Edit, and Read tools, which could be misused if the agent is manipulated by malicious input data.
Sanitization: There are no instructions for escaping or validating user input before it is used in shell commands or prompt generation.

research-grants