youtube-thumbnail-maker

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the agent to PUT bytes to returned upload_url(s) and to "run the curl command provided in the response verbatim", which forces including potentially sensitive presigned upload URLs or tokens verbatim in commands/outputs — an exfiltration risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The workflow ingests outsider-authored free text via WebSearch/WebFetch queries and results (user-provided software names) and then downloads icon URLs from those results (HTML/text may be processed during selection and fetching) before generating the image.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs the agent at runtime to WebSearch for and curl arbitrary icon URLs (including cdn.simpleicons.org/) and then upload those images as medias that are injected into the model generation call, meaning externally fetched content will directly influence the model's outputs.