@howaboua/opencode-glm-gemini-prompt-enhancer

Automatically enhances GLM-4.7 and Gemini models with a rigorous coding assistant system prompt.

What It Does

This plugin prepends a comprehensive system prompt to GLM-4.7 and Gemini models, encouraging:

- Precision - Execute with accuracy, verify claims before stating
- Honesty - Stay grounded in verifiable facts, report actual actions
- Persistence - Try 3 approaches before escalating, complete tasks fully
- Structured workflow - Plan → Execute → Verify → Report
- Error recovery - Quote, diagnose, fix, verify

Installation

Add to your opencode.json:

``json
{
"plugin": ["@howaboua/opencode-glm-gemini-prompt-enhancer"]
}
`

OpenCode automatically installs plugin dependencies at runtime.

Affected Models

- GLM-4.7: Any model with glm-4.7 in its ID (e.g., zai-coding-plan/glm-4.7)
- Gemini: Any model with gemini in its ID

Example Usage

`json
{
"plugin": ["@howaboua/opencode-glm-gemini-prompt-enhancer"],
"model": "gemini-2.0-flash-exp"
}
`

Testing

1. Add to opencode.json` (see Installation above)
2. Run OpenCode with a GLM-4.7 or Gemini model
3. Send a coding task and verify the response follows the structured format

Expected response format:
- Structure: Solution → rationale (if needed) → next steps
- Tone: Direct, concise, skip preambles
- Verification: Claims verified with tools before stating
- Completion: Tasks fully completed before yielding

License

MIT