name	opencode
description	Delegate complex coding tasks, refactoring, and codebase analysis to the OpenCode CLI.
when_to_use	When you need to delegate code writing, bug fixing, test generation, codebase refactoring, or code analysis to OpenCode CLI.
allowed-tools	Bash

OpenCode Integration Skill

This skill guides you in delegating complex software engineering and codebase manipulation tasks to the external OpenCode CLI (opencode).

YOU ARE BOUND BY THIS CONTRACT

When using this skill to run OpenCode, you are delegating execution to a powerful sub-agent. Follow these instructions precisely to ensure clean execution, proper session management, and context sharing.

Technical Specifications & CLI Options

opencode run [message..]: Main command to execute OpenCode with a specific prompt.
--dangerously-skip-permissions: Automatically approves all tool execution permissions (required for non-interactive execution).
-s, --session <session_id>: Resumes a specific session by ID.

Step-by-Step Execution Protocol

Step 1: Directory & Context Verification

Ensure you run the Bash tool in the project root directory (which is your current working directory).
OpenCode executes in a separate process and does not share your in-memory conversation history. You must pass all target file paths (relative to project root), logs, and database schemas explicitly in the prompt.

Step 2: Session ID Management (In-Memory / Task-Scoped Only)

Session IDs must only be reused within the context of the same active task. Do not share sessions across different tasks.

Initial Run: Start the first run of the task without a session ID:
```
opencode run "$ARGUMENTS" --dangerously-skip-permissions
```
Capture Session ID: Inspect the output of the command. Look for the printed session ID. Keep this ID in your memory/conversation history for the duration of this task.
Follow-up Runs (Same Task Only): If you need to make consecutive runs or follow-up corrections for the same task, pass the captured session ID:
```
opencode run "Now fix the failing test you just observed." --session <session_id> --dangerously-skip-permissions
```
Task Completion: Once the overall task is finished, discard the session ID. Do not store it to any files.

Step 3: Parse and Summarize Output

Extract the final answer and any diffs/changes made to the files from the stdout.
Summarize the changes and report them back to the calling context.

Troubleshooting & Guardrails

Stuck CLI: Ensure --dangerously-skip-permissions is set to prevent the agent from hanging on interactive prompts.
Git State: OpenCode may modify files directly. Check git status or git diff after execution.

Más de este repositorio

mismo repositorio

color-system

Fize/soloQueue

Generate curated color token systems with dark/light mode support. Pick from 5 battle-tested palettes (Neutral Modern, Linear Dark, Claude Warm, Cursor Code, Lovable Pop) or generate a custom palette from a base color. Output is a ready-to-paste tokens.css block with semantic variables, automatic dark-mode switching, and WCAG AA contrast validation.

2026-07-140

dev-method

Fize/soloQueue

Development methodology selector and executor. Auto-selects the appropriate development method (TDD / BDD / API-First / Security-First / Direct Implementation) based on task context, or use the specified method directly. Supports direct invocation: `/dev-method tdd`, `/dev-method bdd`, etc.

2026-07-110

git-flow

Fize/soloQueue

Git workflow covering branch management, normalized commits, remote sync, with staged operation flows. Supports git commit, push, pull, branch, smart commit splitting, AI author tagging, safe branch operations.

2026-07-110

novel

Fize/soloQueue

通用网络小说创作技能 - 支持对小说项目进行初始化、做大纲/蓝图编排、故事构造与钩子编织、章节正文创作，以及文本级+故事级双重闭环质检。

2026-07-110

brand-strategy-methodology

Fize/soloQueue

End-to-end brand visual identity design methodology — a six-stage workflow from brand detective to system architect. Covers brand soul discovery, visual strategy, creative exploration, system building, multi-touchpoint narration, and guardian delivery.

2026-07-020

commercial-image-prompt

Fize/soloQueue

Commercial and e-commerce image generation methodology — three-layer prompt engineering framework (subject-scene / style-texture / technical-constraints), scenario-based template library, and iterative optimization strategies for Chinese e-commerce and social media platforms.

2026-07-020

OpenCode Integration Skill

This skill guides you in delegating complex software engineering and codebase manipulation tasks to the external OpenCode CLI (opencode).

YOU ARE BOUND BY THIS CONTRACT

Technical Specifications & CLI Options

opencode run [message..]: Main command to execute OpenCode with a specific prompt.

--dangerously-skip-permissions: Automatically approves all tool execution permissions (required for non-interactive execution).

-s, --session <session_id>: Resumes a specific session by ID.

Step-by-Step Execution Protocol

Step 1: Directory & Context Verification

Ensure you run the Bash tool in the project root directory (which is your current working directory).

OpenCode executes in a separate process and does not share your in-memory conversation history. You must pass all target file paths (relative to project root), logs, and database schemas explicitly in the prompt.

Step 2: Session ID Management (In-Memory / Task-Scoped Only)

Session IDs must only be reused within the context of the same active task. Do not share sessions across different tasks.

Initial Run: Start the first run of the task without a session ID:

opencode run "$ARGUMENTS" --dangerously-skip-permissions

Capture Session ID: Inspect the output of the command. Look for the printed session ID. Keep this ID in your memory/conversation history for the duration of this task.

Follow-up Runs (Same Task Only): If you need to make consecutive runs or follow-up corrections for the same task, pass the captured session ID:

opencode run "Now fix the failing test you just observed." --session <session_id> --dangerously-skip-permissions

Task Completion: Once the overall task is finished, discard the session ID. Do not store it to any files.

Step 3: Parse and Summarize Output

Extract the final answer and any diffs/changes made to the files from the stdout.

Summarize the changes and report them back to the calling context.

Troubleshooting & Guardrails

Stuck CLI: Ensure --dangerously-skip-permissions is set to prevent the agent from hanging on interactive prompts.

Git State: OpenCode may modify files directly. Check git status or git diff after execution.