The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data from web pages through the agent-browser tool.
Ingestion points: The agent-browser tool (documented in browser-testing.md) is used to open, read, and interact with external web pages which may contain attacker-controlled content.
Boundary markers: There are no explicit instructions or delimiters defined to separate or ignore potentially malicious instructions embedded in the content retrieved from the browser.
Capability inventory: The skill possesses significant capabilities including writing to the local file system, executing shell commands (npm, vite), and managing cloud hosting configurations via MCP tools.
Sanitization: No sanitization or filtering logic is specified for processing content gathered from external web surfaces before it is presented to the agent's context.
[EXTERNAL_DOWNLOADS]: Fetches the CloudBase JavaScript SDK from a vendor-owned CDN (https://static.cloudbase.net/cloudbase-js-sdk/latest/cloudbase.full.js) and references official NPM packages (@cloudbase/js-sdk, @cloudbase/node-sdk) for platform integration.
[COMMAND_EXECUTION]: Instructs the agent to execute standard engineering commands for building and testing web applications, such as npm run dev, vite build, and tsc --noEmit.

web-development