discord-delete-own-dm-messages
Fail
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructs the agent to programmatically extract the Discord authentication token ('auth token') from the browser's execution environment. This token provides full access to the user's account and session.
- [COMMAND_EXECUTION]: The skill utilizes the
evaluate_scripttool to inject and run arbitrary JavaScript within the security context of the Discord web application, bypassing standard user interface controls. - [DATA_EXFILTRATION]: The skill accesses highly sensitive data, including private message history and active session tokens. Although it specifies that the token should not be returned in logs, the infrastructure for accessing and processing this data is central to the skill's operation.
- [PROMPT_INJECTION]: The skill is designed to read and process content from Discord DMs, which creates a surface for Indirect Prompt Injection. Maliciously crafted messages from third parties could attempt to influence the agent's behavior, a risk acknowledged by the skill's instruction to ignore third-party text.
Recommendations
- AI detected serious security threats
Audit Metadata