The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions include explicit directives to bypass user interaction for sensitive tasks. Specifically, it states 'invoke the revx-auth skill immediately and apply the fix yourself. Do not interrupt the user to ask them to run setup' in the event of authentication failures. This is a behavioral override that reduces user oversight of the agent's actions.
[CREDENTIALS_UNSAFE]: The skill directs the agent to interact with and autonomously 'fix' sensitive authentication resources, including private keys and configuration files located at ~/.config/revolut-x/. It also instructs the agent on how to handle authentication errors like 'No private key found' or 'Invalid signature' by performing background operations on these credentials.
[PRIVILEGE_ESCALATION]: By instructing the agent to 'apply the fix yourself' for authentication and permission-related errors (insecure permissions), the skill encourages the agent to perform administrative or configuration-level tasks autonomously that would normally require user confirmation or manual intervention.
[INDIRECT_PROMPT_INJECTION]: The skill implements a monitoring loop that reads and reports external data (price alerts, indicator triggers, and event history). This represents an attack surface where malicious data embedded in the monitored streams or alert descriptions could influence the agent's behavior if not properly sanitized.
Ingestion points: The agent periodically reads background task output files generated by the revx monitor commands and the revx events history.
Boundary markers: None identified in the skill instructions to separate command output from agent instructions.
Capability inventory: The skill uses the Bash tool with run_in_background: true and the Read tool to monitor status.
Sanitization: No specific sanitization or filtering logic is mentioned for the output read from the monitoring tasks.

revx-monitor