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 agentmods add skills/ayukyo/icode-skill/icode-skillnpx skills add ayukyo/icode-skill --skill icode-skillgit clone --depth 1 https://github.com/ayukyo/icode-skillWrote 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/ayukyo/icode-skill/icode-skill)<a href="https://agentmods.dev/skills/ayukyo/icode-skill/icode-skill"><img src="https://agentmods.dev/badge/skills/ayukyo/icode-skill/icode-skill.svg" alt="Measured on agentmods" 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 | $0.00388 | $0.49210 |
| Opus 5 | $0.00194 | $0.24605 |
| Sonnet 5 | $0.00078 | $0.09842 |
| Haiku 4.5 | $0.00039 | $0.04921 |
Grade A, and why
icode scanned grade A with 1 finding 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 4d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -s ... # 发一个请求，停，等用户推进 How it starts
The opening of the file, as written. The whole thing — 1,048 lines — stays where its author put it; the contents beside it link to each section on GitHub.
版本: v2.18.0
ICode 全流程编码工作流（步骤 0 + 1~6）
端到端编码工作流，将需求到交付拆解为严格步骤，每步可单独调用，方便你自行切换模型。
- 步骤 0（可选）：需求初稿对话，多轮迭代后落档为
00_init.md（含链路图：修改前/后链路 + 改动点，每轮动态更新），独立步骤、不自动串联到步骤1 - 步骤 1~6：拟定计划 → 审查 → 定稿 → 编码 → 复检 → 终审
主流程步骤真源（防误用，唯一真源 =
steps/目录，启动强制 Read）：步骤编号 / 产物文件名 /completed_steps合法值一律以steps/目录实时清单为准（ls steps/*.md完整列出，含主流程与辅助入口 log / doc / limit / status / install / list / bak，及精简全流程入口 fast）——本块仅示意，steps/ 演进后以目录为准，勿依赖写死。/icode start//icode fast//icode plan进入第一步先ls steps/*.md核对，不按"编码→测试→部署"直觉推断
- 当前主流程示意（以
ls steps/*.md为准）：00_init → 01_plan → 02_review → 03_merge → 04_code → 05_deepcheck → 06_audit → 07_readme → 08_patch；不存在03_code/04_test/05_deploy（测试验证在 04_code 子段，部署/回归归 07_readme / 08_patch）- 辅助独立步骤（doc / log / limit / status / install / list / bak）不参与 1
6 推进；fast 为精简全流程（非辅助独立步骤）——参与 16 但各步缩略（见命令表 fast 行）- 强制：产物命名 +
completed_steps写号对照ls steps/*.md实时结果（如入口含log→ 可写"log"），不在清单 → 停下核对，禁止自造产物占位；steps/ 目录与本文档不一致时以 steps/ 目录为准
通用约定（对话语言）
AI 对用户的回复一律使用中文（提示 / 解释 / 报告 / 追问 / 总结 / 决策说明 / 输出行）。工程内容保持原样、不翻译：代码、标识符、产物文件名、命令、日志原文、报错原文、设备输出、配置字段值。仅当用户明确要求英文回复时切换。
适用所有
/icode命令的会话交互与步骤内对用户的询问/报告；产物文件正文遵循既有中文风格撰写。
调用命令
所有输出保存在 .icode_output/.icode_output_N/（N 自动递增）目录下——所有产物统一收纳在 .icode_output/ 父目录内，避免工程根目录堆积大量 .icode_output_* 目录：
| 命令 | 功能 | 创建目录？ |
|---|---|---|
[辅助] /icode help |
帮助：输出使用流程示例 | 否 |
[辅助] /icode install |
MCP 环境检查+一键安装（独立步骤）：跑 mcp/install.sh 扫描所有 mcp/*/install.sh，每个子工程自检环境（venv/Node/npm）并缺啥补啥、注册到当前宿主（默认 Claude Code=~/.claude.json；--client codex|all 可额外注册 Codex）。新 clone 仓库 / 新机器 / CI 初始化时跑一次。不创建工单目录、不写工单 metadata、不参与 1~6 推进（详见 steps/install.md） |
否 |
[入口] /icode log [零散信息...] |
可选入口（日志根因分析）：把"设备/服务日志+模糊症状"转为有对抗验证的根因报告，自动转修复需求 00_init.md 衔接步骤1。先基线检查（git diff/链路图 + 现场运行版本基线门：按证据优先级从日志/版本文件提取现场运行 Hash → 绑定模块仓库 → 按现场 Hash 读历史代码核对根因、与当前 HEAD 演进对照（判定矩阵），防"用当前 HEAD 解释现场日志"，见 steps/log.md「现场运行版本基线门」+ 步骤9.6「版本基线完成门」）再日志侦察，对抗分析防确认偏误。领域无关，本地日志每次调用都新建目录；同 TB 单再次分析（TB 上有新评论/附件）除外——提示"复用旧工单/新建"，复用则重拉最新+增量对抗（见方式D2）；批量 TB 分析（所有"打开/未完成"单）——零散输入含"分析所有TB/全部打开单"类批量意图时触发（见 steps/log.md「批量 TB 分析」段；按真实任务流状态名过滤，不能用 isDone；"所有TB单"默认=打开/未完成两态，其它状态默认不分析除非显式指定；支持 --debug——整批全部入 .icode_output/.debug/ 域、复用只匹配 debug 孪生（含中断半成品复用续跑，见 references/debug_mode.md §12）；--worktree 不与批量叠加）。（详见 steps/log.md）；报告完成自动生成对外简报（log_problem_brief.md，TB 分析时带单号前缀为 <单号>_log_problem_brief.md 如 DEMO-26_log_problem_brief.md；跨领域交付，剔除流程术语，对外表达按统一契约——归因分级/角色澄清/修复状态明确，见 references/external_brief_contract.md + steps/log.md「对外简报」段）。支持 --worktree（opt-in）；支持 --debug（独立孪生不入索引，详见 references/debug_mode.md） |
✅ 每次都新建（同 TB 单复用除外） |
[入口] /icode init [<粗略需求>] |
可选步骤0：多轮对话产出 00_init.md（需求初稿，含链路图：before/after + 改动点，每轮动态更新）。每次调用都新建目录，不复用、不续聊（详见 steps/00_init.md）。支持 --worktree（opt-in）；支持 --debug（独立孪生不入索引，详见 references/debug_mode.md） |
✅ 每次都新建 |
[流程] /icode start <需求> |
全流程（full 模式）：创建/复用目录 → 步骤1→6 串联。步骤2 review 默认 3 轮 + 对抗验证，步骤5 deepcheck 三阶段循环（复用规则见下）。支持 --worktree（opt-in） |
✅ 创建新目录 / 复用 |
[流程] /icode fast <需求> |
精简全流程（fast 模式）：plan → review(1轮无对抗) → merge → code → deepcheck(Reverse 单阶段) → audit。耗时约为全流程 65%，产物结构与 full 对齐（详见 steps/fast.md）。入口打印警告、用户自负其责。支持 --worktree（opt-in） |
✅ 创建新目录 / 复用 |
[流程] /icode plan <需求> |
仅步骤1：拟定项目计划（复用规则见下）。支持 --worktree（opt-in） |
✅ 创建新目录 / 复用 |
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.
- 4d ago First seen · 1,048 lines · 388 tokens per session scan A 478a16bd00e6
icode is a skill published in the GitHub repository ayukyo/icode-skill (3 stars, last pushed yesterday), licensed MIT. It adds 388 tokens to every session and 49,210 once invoked, about $0.0019 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
author-skill
Use when authoring a NEW rsc skill or editing an existing one — scoping it to one job, writing the description that decides whether it ever loads, splitting the body into references/, writing its evals, auditing it against the rubric. NOT building a product feature (that is specify) and NOT designing an agent loop…
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code. Enforces red-green-refactor cycle with strict rules.
writing-plans
Use for complex multi-step tasks. Create a structured plan before writing code. Use when the user asks for a plan, spec, or when the task is too large to implement in one step.
agent-orchestrator-v2
Agent Orchestrator workflow skill. Use this skill when the user needs Meta-skill que orquestra todos os agentes do ecossistema. Scan automatico de skills, match por capacidades, coordenacao de workflows multi-skill e registry management and the operator should preserve the upstream workflow, copied support files, and…
agent-team
Build low-context multi-agent teams with durable handoff, independent review, and user gates. Use when a project needs separate management, execution, and review roles.
self-improvement
Captures lessons and promotes recurring patterns.