The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation in README.md and SKILL.md instructs users to pull and run an unverified third-party Docker image wwwzhouhui569/jimeng-free-api-all. Executing code from unknown third-party sources represents a supply chain risk, as the image's behavior and contents are not audited.
[DATA_EXFILTRATION]: The scripts/handler.ts file programmatically accesses a sensitive file path at ~/.nby-skills/nby-jimeng-api/.env to retrieve a JIMENG_SESSION_ID. This credential is used as a Bearer token in HTTP requests to localhost:8000. Accessing authentication tokens from the home directory and transmitting them over the local network carries risk of credential exposure.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface (Category 8) because user-supplied input is interpolated into prompt templates without protection.
Ingestion points: The topic parameter in the generate function and command-line arguments in the main function within scripts/handler.ts.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when the topic variable is concatenated into the PROMPT_TEMPLATES strings.
Capability inventory: The skill performs network operations via http.request and https.request, and writes generated image data to the file system using writeFileSync (scripts/handler.ts).
Sanitization: No input validation, escaping, or filtering is performed on the topic string before it is used in prompt construction.

nby-xhs-cover