delphi-seat

Warn

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to run the command delphi seat $1 --json. Because the $1 argument is interpolated directly into the shell command without quoting or sanitization, an attacker providing an argument with shell metacharacters (such as ;, |, or &) could execute arbitrary commands in the agent's environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to load and follow instructions from external and dynamic sources, which could be exploited to manipulate the agent.
  • Ingestion points: The agent is directed to read .delphi/PROTOCOL.md and the output from the delphi command into its active context.
  • Boundary markers: No delimiters or safety instructions are provided to separate these external inputs from the agent's primary system instructions.
  • Capability inventory: The agent has permissions to execute shell commands, read local files, and send messages to other agent seats.
  • Sanitization: The skill does not describe any mechanisms for validating the input argument or filtering the content of ingested files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 18, 2026, 04:46 AM
Security Audit — agent-trust-hub — delphi-seat