The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external benchmark result files to generate SVG figures, alt-text descriptions, and captions. This processing of untrusted data represents an indirect prompt injection surface.
Ingestion points: The skill reads machine-readable measurement data from JSON files located in tests/bench/<family>/results/ (referenced in SKILL.md).
Boundary markers: The instructions do not specify explicit boundary markers or "ignore instructions" delimiters when interpolating data into the chart's labels or the generated alt-text, which could allow malicious data to influence agent behavior during chart generation.
Capability inventory: The skill possesses the capability to write files to the local filesystem using fs.writeFileSync (in scripts/chart.mjs) and describes the use of shell commands like rsvg-convert for rasterization.
Sanitization: The renderer in scripts/chart.mjs includes an esc function that performs basic HTML entity encoding (replacing &, <, and >) for strings placed within the SVG XML structure, providing a layer of protection against simple injection attacks in the visual output.

nub-charts