The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the opencode CLI tool and various slash commands (/sessions, /agents, /models) to manage coding sessions and model configurations.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting content generated by the external Opencode tool.
Ingestion points: Output from Opencode, such as task plans and clarification questions, is read by the agent (found in SKILL.md and assets/operator-prompts.md).
Boundary markers: The instructions do not define explicit delimiters to isolate data from the external tool.
Capability inventory: The agent can execute commands through the opencode interface and its subcommands (documented in references/command-cheatsheet.md).
Sanitization: No specific sanitization or validation of the tool's output is performed before processing.
[SAFE]: No evidence of malicious behavior, credential harvesting, or unauthorized data exfiltration was found. The skill includes user-in-the-loop checkpoints for authentication and session management.

opencode-controller