The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Node.js scripts to execute a browser controller tool (agent-browser) via child_process.execFileSync. This is the standard mechanism for the skill to perform its intended task of web scraping.
[EXTERNAL_DOWNLOADS]: Data is fetched from several major web novel platforms, including Qidian, Fanqie, JJWXC, Qimao, and Ciweimao. These connections are legitimate and necessary for the scanner's operation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text content from external novel websites, such as plot summaries and user-generated tags, which could potentially contain malicious prompts intended to deceive the AI agent.
Ingestion points: Data collection logic in scripts/ciweimao-rank-scraper.js, scripts/fanqie-rank-scraper.js, scripts/jjwxc-rank-scraper.js, scripts/qidian-rank-scraper.js, and scripts/qimao-rank-scraper.js.
Boundary markers: The scraped data is converted into structured Markdown by scripts/aggregate-rank.js before being presented to the agent.
Capability inventory: The skill allows the agent to run Node.js scripts, perform network requests, and write results to the filesystem.
Sanitization: The aggregation script performs basic text cleaning, regex matching, and numerical parsing on the ingested data.
[DYNAMIC_EXECUTION]: The scraper scripts use JavaScript evaluation within the context of a browser (via the Chrome DevTools Protocol) to interact with and extract data from target websites. This execution is confined to the browser instance and does not affect the host system.

story-long-scan