opencode-agent

Warn

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary shell commands on a remote server using the opencode binary. This includes both the predefined setup commands and any arbitrary command passed within a task string.
  • [REMOTE_CODE_EXECUTION]: The primary purpose of the skill is to delegate coding tasks to a remote server, which inherently involves the execution of generated code on that system.
  • [EXTERNAL_DOWNLOADS]: The skill depends on the presence of an external binary named opencode on the local system, which is not provided within the skill itself.
  • [CREDENTIALS_UNSAFE]: The skill requires printing the environment variables OPENCODE_SERVER_URL and OPENCODE_SERVER_DIR to the console, which may expose sensitive network or infrastructure details.
  • [PROMPT_INJECTION]: The skill accepts task descriptions from the user or other data sources and interpolates them directly into shell commands without sanitization or boundary markers, creating a surface for indirect prompt injection.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 12, 2026, 11:39 PM