The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The file scripts/ensure_uv.py performs remote script execution by downloading a shell script from https://astral.sh/uv/install.sh and piping it directly into a subprocess shell. The logic uses urllib.request.urlopen to fetch the script content and subprocess.run(["sh"], input=script) to execute it. This pattern allows an external source to execute arbitrary code on the host system.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality. In addition to the installer script, the SKILL.md instructions and various validation scripts (e.g., validate_configuration.py) utilize subprocess.run to call system commands and project-specific tools.
[DYNAMIC_EXECUTION]: The skill uses dynamic execution to manage its environment. Specifically, scripts/ensure_uv.py fetches and runs code at runtime that is not bundled with the skill package, creating a dynamic execution path from a remote source.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting and processing untrusted data through its validation scripts.
Ingestion points: Data is ingested via CSV files in the assets/ directory, such as config-schema-catalog.csv and naming-convention-catalog.csv.
Boundary markers: There are no boundary markers or instructions to explicitly ignore embedded instructions within the processed data.
Capability inventory: The skill has significant capabilities, including network access (via urllib) and command execution (via subprocess).
Sanitization: While scripts/configuration/validate_configuration.py performs basic type and enum validation, it does not include sanitization to prevent the interpolation of malicious instructions into the agent's context.

engineering-quality