The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill uses eval() in Step 1 of SKILL.md to parse configuration data stored as a JS-escaped string inside the IDE's main extension bundle. This method of unescaping strings is risky as it evaluates the string content in the current process context.
[COMMAND_EXECUTION]: Several shell commands are used to build the MCP bundle (npm run build:webpack) and to execute internal Node.js scripts that modify the IDE's application resources and configuration, such as apply-to-ide.mjs and patch-tool-timeout.mjs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes the content of an external application file (index.js). The use of dynamic execution on this data provides a surface for potential code injection if the source file is compromised.
Ingestion points: The IDE extension bundle Contents/Resources/app/extensions/genie/out/extension/index.js is read to extract embedded configuration.
Boundary markers: None identified; the skill relies on locating specific string anchors (JSON.parse(') to determine which portion of the code to evaluate.
Capability inventory: Includes dynamic code evaluation via eval(), file system writes to the /Applications directory, and shell command execution via Node.js and npm.
Sanitization: There is no validation or sanitization performed on the data slice before it is passed to the eval() function.
[EXTERNAL_DOWNLOADS]: The documentation rendering tool (scripts/render-html.mjs) performs a download of the marked library from the official npm registry.

codebuddy-ide-mcp-upgrade