The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/render-elk.mjs uses child_process.execSync to perform PNG conversions using the macOS sips utility. The command string is constructed by concatenating filenames retrieved directly from the filesystem via readdirSync.
Evidence: Filenames are wrapped in double quotes in the command string: execSync(sips -s format png "${join(outDir, svgName)}" --out "${join(outDir, pngName)}" 2>/dev/null);.
Risk: If a file in the target directory has a name containing shell metacharacters (e.g., ";touch RCE;".json), the shell may interpret the name as separate commands, leading to arbitrary code execution.
[DYNAMIC_EXECUTION]: The script employs dynamic module loading to import its primary dependency.
Evidence: const require = createRequire(process.cwd() + '/'); const ELK = require('elkjs'); in scripts/render-elk.mjs.
Risk: This pattern loads the elkjs library from the user's current working directory rather than a fixed path. This allows the execution of code from the project environment, which is a standard but sensitive operation.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a data processing pipeline that ingests external, potentially untrusted JSON data to generate visual diagrams.
Ingestion points: scripts/render-elk.mjs reads and parses .json files from the filesystem.
Boundary markers: The script does not implement specific boundary markers or instructions to ignore embedded commands within the JSON data, though it is intended for structured data.
Capability inventory: The script possesses file read/write capabilities (fs.readFileSync, fs.writeFileSync) and shell execution capabilities (execSync).
Sanitization: The script uses an escapeXml function to sanitize text labels (label, title, subtitle) before injecting them into the SVG output, which helps mitigate XML/SVG injection risks.

diagrams