The Agent Skills Directory

[SAFE]: The skill provides high-quality instructional content, including agency-specific guidelines, budget templates, and writing guides. All analyzed files are standard Markdown documents with no malicious code, base64 obfuscation, or persistence mechanisms.
[COMMAND_EXECUTION]: The documentation includes example shell commands using the Bash tool to run the uv package runner. These commands (e.g., uv run python scripts/generate_schematic.py) are provided as optional developer tools for generating project timelines and workflow diagrams.
[EXTERNAL_DOWNLOADS]: The skill references authoritative government domains such as nsf.gov, nih.gov, energy.gov, and nstc.gov.tw. It also mentions community-supported LaTeX templates on GitHub, which are standard resources for academic formatting and do not involve automated execution of untrusted remote code.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-provided research descriptions to generate proposal content. While the agent has access to powerful tools, the risk is inherent to the assistant role rather than a flaw in the skill. 1. Ingestion points: User research abstracts and technical narratives processed in SKILL.md workflows. 2. Boundary markers: Absent. 3. Capability inventory: Read, Write, Edit, and Bash tools available in the agent host. 4. Sanitization: Absent.

research-grants