The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local diagnostic scripts and the moviepilot doctor command to gather system information. These commands are constructed safely without shell execution of user-supplied input.
[DATA_EXFILTRATION]: While the skill sends logs and diagnostics to GitHub repositories, this is its primary stated purpose. It includes comprehensive redaction logic in feedback_issue_common.py to strip sensitive data—including API keys, authorization tokens, cookies, email addresses, and IP addresses—before any data is sent externally or displayed to the user.
[CREDENTIALS_UNSAFE]: The skill manages GitHub authentication by reading tokens from the application's secure configuration or environment variables. It explicitly instructs the agent not to ask for or echo tokens in the chat, following security best practices.
[INDIRECT_PROMPT_INJECTION]: The skill includes explicit instructions to treat user-provided logs and text as untrusted data, specifically warning the agent to ignore any instructions embedded within them to prevent manipulation of the target repository or issue content.

feedback-issue