The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates a robust security model to handle untrusted data.
Ingestion points: The skill reads project source code, console logs, and tool outputs during development and auditing (SKILL.md, references/audit.md).
Boundary markers: Present; the 'Security Model' section in SKILL.md explicitly defines the boundary between data and instructions, warning the agent to treat external strings as values to be rendered rather than directives.
Capability inventory: The skill utilizes shell commands for package installation and codebase searching via standard tools (SKILL.md, references/audit.md).
Sanitization: Present; the skill mandates HTML escaping for untrusted data interpolated into tooltip formatters (SKILL.md).
[EXTERNAL_DOWNLOADS]: The provided vanilla JavaScript example loads the ECharts library from JSDelivr, a well-known CDN. The implementation follows security best practices by including a Subresource Integrity (SRI) hash to ensure the integrity and authenticity of the remote script.
[COMMAND_EXECUTION]: The skill provides auditing instructions that involve using standard command-line utilities (such as rg, perl, and cat) to analyze project source code for ECharts usage patterns.
[SAFE]: The instructions include defensive programming requirements, specifically recommending the use of richText mode or manual escaping to prevent Cross-Site Scripting (XSS) vulnerabilities in dynamic tooltips.

echarts