What is invoke-opencode-acp CLAUDE.md?

A set of instructions for using Claude Code with OpenCode, another coding assistant that can handle delegated tasks. It covers when to delegate work, how to use skills, and the commands involved.

Deciding when to hand work to OpenCode, running OpenCode from the command line, tracking multi-step tasks, and following repository-specific rules.

It provides a consistent process for larger coding tasks, reviews, refactors, and Git work. It also explains the trade-off that delegated work uses a different model and returns only a summary.

invoke-opencode-acp CLAUDE.md — Instructions file by phonowell

Install

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.

agentmods

npx agentmods add instructions/phonowell/invoke-opencode-acp/claude-md

Clone the repo

git clone --depth 1 https://github.com/phonowell/invoke-opencode-acp

Per session 727 This file is loaded in full into every session.

When invoked 727 The same file — it is already loaded in full.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

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	$0.00727	$0.00727
Opus 5	$0.00364	$0.00364
Sonnet 5	$0.00145	$0.00145
Haiku 4.5	$0.00073	$0.00073

Measured 2d ago against content hash 86fe269055af, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

invoke-opencode-acp CLAUDE.md 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 2d ago.

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.

CLAUDE.md · 62 lines

What it actually says

invoke-opencode-acp

Claude Code skill · 委托复杂任务至 OpenCode 子代理 · 节省 50-90% token

关键约束

精简冗余：机器可读 > 人类可读 · 符号/缩写/列表 · 删助词 冲突信代码：代码为真 · 文档描述与实现冲突时信代码 客观诚实：不主观评价 · 不因用户情绪转移立场 · 不编造事实 · 立刻暴露不确定信息 触发条件：≥2 文件修改 · 重构 · 批量操作 · 代码审查 · 多步推理 · git 操作避免：单文件快速编辑 · 递归调用 OpenCode Token 机制：委托至 OpenCode → 使用替代模型（GLM-4.7/Qwen）→ 仅摘要入主对话 · Tradeoff: 能力下降 TodoWrite：≥3 步骤任务必须建 todo · 实时更新状态 · 完成立即标记 子任务模型：Task 工具可用 model: "haiku" 节省成本（简单任务）

Skill 使用

调用后必须等待完成再执行其他操作 · 避免立即自行实现

技术栈

Claude Code CLI · OpenCode CLI（npm install -g opencode）· Node.js

核心命令

交互式（>30m）：

node ~/.claude/skills/invoke-opencode-acp/acp_client.cjs "$PWD" "任务描述" -o /tmp/output.txt

一次性（<30m）：

node ~/.claude/skills/invoke-opencode-acp/acp_client.cjs "$PWD" "任务描述" -o /tmp/output.txt -t 300

参数：-o FILE 必需 · -t SECONDS 默认 1800 · -v 详细模式→stderr

工作流

交互式检查循环：

Bash(..., run_in_background=True) → task_id
TaskOutput(task_id, block=True, timeout=300000) → 等待 5 分钟
completed → Read 输出文件 → 返回结果
running → AskUserQuestion 是否继续 → 继续则回到 2
终止 → KillShell(task_id) 自动清理

目录结构

skills/invoke-opencode-acp/
  SKILL.md
  acp_client.cjs
tests/
  test_acp_client.js

代码规范

≥5 处非空断言立即重构类型架构（🚫 eslint-disable 批量压制）

输出格式

禁预告文字 · 状态用符号 ✓/✗/→ · 一次性批量 Edit · 数据优先 · 直达结论 · 工具间隔零输出 · 错误格式 ✗ {位置}:{类型} · 代码块零注释 · ≥2 条用列表 · 路径缩写（. 项目根 · ~ 主目录）· 禁总结性重复 · 进度 {当前}/{总数} · 提问直入

Changes

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.

2d ago First seen · 62 lines · 727 tokens per session scan A 86fe269055af

Subscribe to this mod's changes

invoke-opencode-acp CLAUDE.md is an instructions file published in the GitHub repository phonowell/invoke-opencode-acp (9 stars, last pushed 7mo ago), licensed MIT. It adds 727 tokens to every session, about $0.0036 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 2026-08-31.

Building agents agent-delegation claude-code claude-skills cost-saving opencode token-optimization

lean-ctx AGENTS.md

AGENTS.md instructions for yvgude/lean-ctx, covering context engineering layer, mandatory multi-agent delivery, integration mode: replace, subagent compatibility (cursor / claude code) and cli commands (optimized shell, lower overhead).

yvgude/lean-ctx · 2,223 tokens

lean-ctx copilot-instructions.md

Copilot instructions for yvgude/lean-ctx, a project described as: Control what your AI can see. LeanCTX (Lean Context) is the context intelligence layer for AI agents — one local Rust binary that decides what they read, remembers what they learn, guards what they touch, and proves what they save. 60–90% fewer tokens…

yvgude/lean-ctx · 253 tokens

invoke-opencode-acp CLAUDE.md