The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute multiple Node.js scripts located within the skill's directory (e.g., qidian-rank-scraper.js, fanqie-rank-scraper.js, scan-analyze.js) to collect and process data.
[DYNAMIC_EXECUTION]: The utility file scripts/cdp-utils.js implements a function resolveWindowsAgentBrowser that dynamically discovers, reads, and parses command-line shim files from the system PATH to resolve executable paths. Furthermore, the scraper scripts use browser-based eval to execute dynamically constructed JavaScript strings for data extraction from web pages.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it ingests and analyzes content from multiple external novel platforms. • Ingestion points: Scraper scripts fetch book titles, author names, and descriptions from qidian.com, fanqienovel.com, jjwxc.net, and qimao.com into the agent's context. • Boundary markers: The instructions do not provide explicit delimiters or safety warnings to the agent for handling the untrusted scraped content. • Capability inventory: The skill has the ability to execute local shell commands via execFileSync and write to the local filesystem. • Sanitization: While the skill performs basic formatting and truncation of descriptions, it does not sanitize retrieved data for potential malicious prompt instructions.
[EXTERNAL_DOWNLOADS]: The skill performs automated network operations using https.get, fetch, and XMLHttpRequest to download content from third-party domains such as qidian.com, fanqienovel.com, jjwxc.net, and qimao.com.

moshu-scan