opencode-deepseek-sidecar

Delegates bounded tasks to DeepSeek-backed OpenCode worker agents via a sidecar script.

21|3|Updated Jun 1, 2026
One-click install
npx skills add https://github.com/Zedong-Liu/codex-deepseek-sidecar --skill opencode-deepseek-sidecar-zedong-liu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: opencode-deepseek-sidecar
Source: https://github.com/Zedong-Liu/codex-deepseek-sidecar/tree/main/.opencode/skills/opencode-deepseek-sidecar
Command: npx skills add https://github.com/Zedong-Liu/codex-deepseek-sidecar --skill opencode-deepseek-sidecar-zedong-liu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running long, token-heavy subtasks like log analysis, broad code exploration, or extended test runs on expensive primary models consumes significant budget. This Skill lets OpenCode offload those bounded tasks to cheaper DeepSeek worker agents while the main agent keeps planning and synthesizing. ## Core Features & Use Cases - Sidecar Task Delegation: Launch a self-contained OpenCode worker session backed by DeepSeek models through the repository's adapter script. - Flexible Credential Handling: Pass the DeepSeek API key via environment variable or read it from a private local file with --api-key-file. - Optional Local Proxy: Start the built-in Python responses proxy when the Pro/Beta transport path requires it. - Use Case: Ask OpenCode to analyze a large CI failure log with a DeepSeek sidecar, returning a root-cause report while the main agent continues reviewing code. ## Quick Start Ask OpenCode to use the opencode-deepseek-sidecar skill to run a bounded task such as analyzing a CI log with a DeepSeek worker and report the root cause.

Frequently Asked Questions about opencode-deepseek-sidecar

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a DeepSeek sidecar task from OpenCode?▼

Run the .opencode/scripts/deepseek-sidecar script from the repository root with --cd pointing at the project, a --model such as deepseek/deepseek-v4-pro, and a task string describing the bounded task, expected output, and constraints.

How do I pass a DeepSeek API key to the sidecar script?▼

Set the DEEPSEEK_API_KEY environment variable before launching, or use the --api-key-file flag to read the key from a private local file such as ~/.codex/deepseek-sidecar.key.

Does the OpenCode sidecar use Codex profiles or the Responses API?▼

No. The OpenCode adapter uses OpenAI-compatible chat completions and does not rely on Codex profiles or the Codex Responses translation path used by the Codex variant of this skill.

When do I need the local DeepSeek responses proxy?▼

The built-in Python proxy is only needed for Pro/Beta model paths that require it. Start it with scripts/deepseek-responses-proxy and keep it running before launching the sidecar task.

What kinds of tasks should I delegate to a DeepSeek sidecar?▼

Delegate bounded, well-scoped tasks such as long test runs, log analysis, wide code exploration, or independent reviews. Keep planning, judgment, and synthesis with the primary agent, and define expected output and constraints in the task string.