The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a shell command using a local Python helper (scripts/buffer_post.py) with user-supplied data interpolated directly into the command arguments (e.g., --text "<copy>"). This pattern is vulnerable to command injection if the user content contains shell-metacharacters (like backticks or semicolons) that could allow an attacker to execute arbitrary system commands.
[EXTERNAL_DOWNLOADS]: The Python helper script communicates with the well-known Buffer API service at api.buffer.com. This is legitimate functionality for the skill's purpose, but it involves external network communication.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user data (the article content) which is then passed into a shell command. While the 'Preview & Confirm' step provides some oversight, it may not prevent sophisticated injection attacks targeting the underlying system shell.
[SAFE]: The skill uses a .secrets.env file to manage the BUFFER_ACCESS_TOKEN. This is a standard and safe practice for local development, and the author correctly advises users to gitignore this file and avoid sharing its contents.
[SAFE]: The implementation of a mandatory human-in-the-loop confirmation step before any content is posted live is a critical security control that reduces the risk of accidental or malicious automated posting.

social-publish-article