The Agent Skills Directory

[SAFE]: The skill is a standard text-processing utility. It analyzes user-provided script files against a local database of keywords to identify potential platform violations. All operations are performed locally within the project environment.
[COMMAND_EXECUTION]: The skill includes a Node.js script (scripts/scan.mjs) used for scanning text. This script only uses built-in Node.js modules (node:fs, node:crypto, node:path, node:url) to read local files and calculate SHA-256 hashes. It does not execute shell commands, perform network requests, or use dangerous functions like eval().
[DATA_EXPOSURE]: File access is restricted to the specific project structure defined in the instructions. The scanning script reads from references/ and episodes/ directories and writes a JSON report to the same episode folder. It does not attempt to access system credentials, environment variables, or files outside the project scope.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input (user-generated video scripts). However, the processing logic is limited to static keyword matching and normalization. The script includes a sanitization function (normalize) that strips non-alphanumeric and non-Han characters, reducing the risk of malicious payload execution during the scanning phase.
[OBFUSCATION]: No obfuscated code, hidden URLs, or suspicious character encoding techniques were detected in the instructions or scripts.

laohan-weigui