The Agent Skills Directory

[COMMAND_EXECUTION]: The execution script scripts/run.py utilizes os.execv to re-execute the script within a dedicated virtual environment (.runners-venv). This is a common and safe implementation pattern for CLI tools to ensure dependency isolation and does not represent a malicious privilege escalation or execution attempt.
[INDIRECT_PROMPT_INJECTION]: The skill ingests content from external files, such as captions.md, which are likely generated by other agents. This creates an attack surface where malicious instructions could be embedded in the content destined for social media platforms. The skill robustly mitigates this via its design: it enforces a dry-run by default, provides a preview of the extracted text, and requires explicit user confirmation for each platform before any data is transmitted externally.

post-publisher