The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by ingesting untrusted job invitation text pasted by the user. However, the analysis shows that the skill possesses no dangerous capabilities such as network access, file writing, or command execution. Consequently, the risk is negligible as the agent only generates text drafts based on the input.
[DATA_EXFILTRATION]: The skill is designed to read a local configuration file (upwork-profile.md or files within ~/.claude/) to retrieve user profile information. This is a legitimate functional feature for personalizing response drafts and does not involve any exfiltration or transmission of data to external services.

upwork-invitation-responder