The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to invoke gnuplot for generating PNG/SVG/sixel charts. The implementation is secure as it uses a fixed executable list (disabling shell interpolation) and provides script content via standard input. Strings derived from input data (such as titles and labels) are sanitized using json.dumps, which provides appropriate escaping for gnuplot's double-quoted string syntax, preventing command breakout.
[INDIRECT_PROMPT_INJECTION]: The skill processes metrics data, including user-controlled tags and metric names, which are incorporated into labels and legends. While this represents an ingestion surface for untrusted data, the skill lacks dangerous capabilities such as network access or unrestricted file writes, and the output is intended for display in the agent transcript or terminal, posing no risk to the underlying system.
[DATA_EXFILTRATION]: No network operations were detected. The skill operates entirely on local JSON input and limits file operations to writing generated charts to temporary files or user-specified paths.
[OBFUSCATION]: The source code is transparent and well-documented. No techniques such as Base64-encoded instructions, zero-width characters, or homoglyph attacks were found.
[PRIVILEGE_ESCALATION]: The skill does not attempt to acquire elevated privileges and operates within the standard user environment.

metrics-chart