The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions require the installation of third-party software from untrusted sources.
SKILL.md instructs the user to execute npm install -g fireworks-tech-graph and fireworks-cli --version.
references/concept-diagrams.md directs the user to clone a repository from https://github.com/NousResearch/concept-diagrams.git and run its scripts.
These sources are not recognized as trusted vendors, and the fireworks-tech-graph package name does not correspond to a widely known or verifiable tool.
[DYNAMIC_EXECUTION]: The skill executes Python scripts from computed relative paths.
scripts/render_diagram.py contains logic to locate and run generate_pdf.py at ../../../scripts/generate_pdf.py using subprocess.run.
Executing scripts from arbitrary relative locations outside the skill's package is a potential security risk if the environment contains unexpected files.
[COMMAND_EXECUTION]: The skill invokes external command-line tools with arguments derived from user-supplied data.
scripts/render_diagram.py uses subprocess.run to call fireworks-cli with various parameters derived from the input JSON spec.
SKILL.md includes Python code that executes which fireworks-cli or where fireworks-cli.exe to detect the presence of the tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted JSON specifications which can contain text that is later rendered into diagrams.
Ingestion points: The JSON input file processed by scripts/render_diagram.py.
Boundary markers: The skill lack specific boundary delimiters or instructions to ignore embedded content.
Capability inventory: The skill can execute shell commands and Python scripts.
Sanitization: While _html_escape is used in the built-in renderer, the CLI path relies on the external tool's sanitization logic.

nextreme-diagrams