Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add znlgis/opengis-skills --skill my-opencode-deepseek-configgit clone --depth 1 https://github.com/znlgis/opengis-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/znlgis/opengis-skills/my-opencode-deepseek-config)<a href="https://agentmods.dev/skills/znlgis/opengis-skills/my-opencode-deepseek-config"><img src="https://agentmods.dev/badge/skills/znlgis/opengis-skills/my-opencode-deepseek-config/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/znlgis/opengis-skills/my-opencode-deepseek-config"><img src="https://agentmods.dev/badge/skills/znlgis/opengis-skills/my-opencode-deepseek-config.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00092 | $0.02734 |
| Opus 5 | $0.00046 | $0.01367 |
| Sonnet 5 | $0.00018 | $0.00547 |
| Haiku 4.5 | $0.00009 | $0.00273 |
Grade A, and why
my-opencode-deepseek-config scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.
项目地址： https://github.com/znlgis/my-opencode-deepseek-config
许可证： MIT License（最新 tag：v4.0.0）
文档： https://github.com/znlgis/my-opencode-deepseek-config/blob/main/README.md（中英双语）
前置条件： OpenCode ≥ v1.18.x（DeepSeek provider 内置）、DeepSeek API Key（https://platform.deepseek.com/api_keys）
概述
OpenCode 多 Agent 框架下 DeepSeek V4 模型族（Pro + Flash + Flash-Vision）的最优纯配置方案——零额外依赖，全部能力由 opencode.jsonc + agents/*.md + skills/*/SKILL.md + AGENTS.md 实现。核心理念：Token 效率优先，用最小的上下文成本达到最好的开发效果。
当前配置概览：
- 默认主 Agent
orchestrator（意图门控 + 模型感知路由 + 后备链）；第二 primarysolo（单模型内联执行、零委派） - 3 模型矩阵：
deepseek/deepseek-v4-pro（深度推理/审查/重型实现）、deepseek/deepseek-v4-flash（路由/规划/常规执行，provider 层关 thinking + temperature 0）、deepseek/deepseek-v4-flash-vision-exp（多模态） - 12 个 Agent、25 个技能（按需加载）、17 个 slash 命令、
subagent_depth: 3、share: "disabled" - 上下文双引擎：内置 compaction（opencode.jsonc，自动触发 + prune 旧工具输出）+ DCP 插件（dcp.jsonc，主动去重 + 按模型成本分级压缩阈值）
环境要求： OpenCode ≥ v1.18.x；后台子智能体委派需环境变量 OPENCODE_EXPERIMENTAL_BACKGROUND_SUBAGENTS=true（未设置时后台委派报错，应退回前台串行）。
安装部署
git clone https://github.com/znlgis/my-opencode-deepseek-config.git
方式一（推荐）：环境变量指向 opencode/ 子目录
# Windows 永久生效
[Environment]::SetEnvironmentVariable("OPENCODE_CONFIG_DIR", "D:\path\to\my-opencode-deepseek-config\opencode", "User")
# Linux / macOS
export OPENCODE_CONFIG_DIR="$HOME/path/to/my-opencode-deepseek-config/opencode"
方式二：符号链接到 ~/.config/opencode（Windows 需管理员 New-Item -ItemType SymbolicLink；类 Unix ln -s .../opencode ~/.config/opencode）
API Key： TUI 内 /connect → DeepSeek → 粘贴 Key（自动存 ~/.local/share/opencode/auth.json），或环境变量 DEEPSEEK_API_KEY。
验证： /models 显示 deepseek/deepseek-v4-pro；Agent 列表含 orchestrator/planner/deep-worker 等 12 个。
Agent 结构（12 个）
| Agent | 模型档 | 权限 | 作用 |
|---|---|---|---|
orchestrator |
flash · thinking 关 | 读写 | 默认入口：意图门控 + 路由 + 后备链 |
solo |
pro · 默认 high | 读写 | 单模型内联执行器，permission.task: "*": "deny" 零委派 |
planner |
flash + reasoningEffort: low |
读写 | 规划、架构、拆解任务 |
deep-worker |
pro | 读写 | 重型实现、多文件改动（禁研究/禁委托，带拒绝契约） |
light-orchestrator |
flash + low | 读写 | 轻量任务、单文件编辑 |
oracle |
pro | 只读 | 根因分析、深度理解代码 |
reviewer |
pro | 只读 | 单遍代码审查（证据门控） |
explore / librarian |
flash · thinking 关 | 只读 | 代码库搜索 / 文档检索（librarian 无任何 bash 权限） |
consultant / ui-builder |
flash · thinking 关 | 读写 | 方案建议 / 前端 UI |
vision |
flash-vision-exp | 读写 | 图像/截图/图表/UI 稿理解 |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 146 lines · 92 tokens per session scan A 451bbdf1019a
my-opencode-deepseek-config is a skill published in the GitHub repository znlgis/opengis-skills (60 stars, last pushed yesterday), licensed MIT. It adds 92 tokens to every session and 2,734 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-10.
Other skills, from other repositories
gh-cli
Patterns for invoking the GitHub CLI (gh v2.100.0+) from agents. Use when the task mentions GitHub, gh, pull requests/PRs, issues, releases, gists, Actions/workflow runs, forks, repo cloning, reviews, or you need exact gh commands. Covers pagination, repo targeting, search vs list, discussions, projects, rulesets…
code-review
A single-pass review method for examining a code change, branch, or pull request. A pull request is a proposed change for review before it is merged into a shared codebase.
spec-workflow
Run a lightweight, spec-driven change workflow — proposal (WHY/WHAT), specs (WHAT), design (HOW), tasks checklist, then archive. Use when the task is a non-trivial feature or behavior change that benefits from durable, git-tracked planning artifacts, or when the request mentions "spec", "proposal", "propose a change"…
git-master
Advanced Git operations — rebase, squash, fixup, blame, bisect, reflog, code archaeology (git log -S/-G), worktrees. Use when the task mentions rebase, squash, bisect, blame, reflog, fixup, finding deleted code, tracing commit history, cleaning up a branch, or recovering lost work.
simplify
Behavior-preserving code simplification — reduce complexity without changing what the code does. Use when the task mentions "simplify", "reduce complexity", "too clever", "hard to read", "reduce nesting", or after a feature lands and the code needs polishing. Oracle analyzes (read-only), light-orchestrator applies the…
verify-with-docs
Verify a specific or fast-moving library, framework, or API against its current documentation before writing code — retrieval-first, never from memory. Use when implementing against a named dependency (especially one that changes fast or whose exact signatures matter), when unsure of an API's current shape, or when…