The Agent Skills Directory

[DATA_EXFILTRATION]: The skill facilitates uploading HTML files to an external service (ARTIFACT_VIEWER_URL). While this is the primary purpose of the skill, the documentation explicitly warns against publishing sensitive data such as private keys or credentials. The implementation in scripts/publish_artifact.py relies on user-provided environment variables for the target URL and authentication token, which is a standard and safe practice for secret management.
[COMMAND_EXECUTION]: The SKILL.md file provides example usage commands for running the Python publisher script. The script itself is a standard utility that reads a local file and performs an HTTPS POST request using the built-in urllib library. No arbitrary shell execution or unsafe command construction was detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided HTML files. There is a potential attack surface where instructions could be embedded within the HTML to influence the agent during the 'live verification' phase mentioned in the workflow. However, the documentation includes a remediation step (the 'Completion checklist') requiring manual inspection of the rendered artifact, which serves as a human review checkpoint.

artifact-viewer-publishing