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 strzhao/autopilot --skill autopilot-codexgit clone --depth 1 https://github.com/strzhao/autopilotWrote 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/strzhao/autopilot/autopilot-codex)<a href="https://agentmods.dev/skills/strzhao/autopilot/autopilot-codex"><img src="https://agentmods.dev/badge/skills/strzhao/autopilot/autopilot-codex/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/strzhao/autopilot/autopilot-codex"><img src="https://agentmods.dev/badge/skills/strzhao/autopilot/autopilot-codex.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 | $0.00077 | $0.01540 |
| Opus 5.5 | $0.00031 | $0.00616 |
| Sonnet 5.5 | $0.00015 | $0.00308 |
| Haiku 4.5 | $0.00008 | $0.00154 |
Grade A, and why
autopilot-codex scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Autopilot for Codex
这是本仓库 autopilot 的 Codex 兼容入口。目标是在 Codex 中恢复与 Claude autopilot 等价的 phase-state runtime,而不是停留在“方法论兼容”。
基线文件
把以下文件当作只读基线:
plugins/autopilot/skills/autopilot/SKILL.mdplugins/autopilot/skills/autopilot/references/
Claude 专有机制只作为行为 oracle,不直接把运行时写回 .claude/。
Codex 兼容边界
- 运行时状态文件:
.codex/autopilot.local.md - 不要写入
.claude/autopilot.local.md - 可以依赖 repo-local
.codex/hooks.json的Stopruntime .autopilot/可作为只读知识库- 不恢复历史 plugin-sync、bridge、watcher
Runtime Helpers
先解析状态脚本路径:
AUTOPILOT_STATE_SCRIPT="$(git rev-parse --show-toplevel 2>/dev/null)/codex/plugins/autopilot-codex/assets/scripts/autopilot_state.py"
if [ ! -f "$AUTOPILOT_STATE_SCRIPT" ]; then
AUTOPILOT_STATE_SCRIPT="${CODEX_HOME:-$HOME/.codex}/plugins/cache/string-codex-plugins/autopilot-codex/local/assets/scripts/autopilot_state.py"
fi
统一使用它管理状态:
python3 "$AUTOPILOT_STATE_SCRIPT" start --goal "<GOAL>"
python3 "$AUTOPILOT_STATE_SCRIPT" approve [--feedback "<TEXT>"]
python3 "$AUTOPILOT_STATE_SCRIPT" revise --feedback "<TEXT>"
python3 "$AUTOPILOT_STATE_SCRIPT" status [--json]
python3 "$AUTOPILOT_STATE_SCRIPT" cancel [--reason "<TEXT>"]
用户意图映射
把当前用户请求视为以下两类之一:
- 启动新流程
- 控制已有流程:
approve/revise <反馈>/status/cancel
如果请求更像提交或诊断:
autopilot commit->$autopilot-commit-codexautopilot doctor->$autopilot-doctor-codex
状态写回规则
- 每次改
phase、gate、retry_count、qa_scope、updated_at,先写回.codex/autopilot.local.md - design 必须写完
## 目标、## 设计文档、## 实现计划、## 验证方案 - implement 必须把红队产物写回
## 红队验收测试 - qa / auto-fix 必须把每轮证据追加到
## QA 报告,不要覆盖旧轮次 - merge 结束前把最终摘要写回状态文件,再设
phase=done
工作流
1. 启动或恢复
- 新目标:运行
start,然后继续当前 phase - 控制命令:运行对应子命令;如果 phase 被推进到活跃阶段(含
auto-fix、merge),在同一轮继续推进 - 如果没有状态文件,
approve/revise/status/cancel直接说明没有活跃 autopilot
2. Design
- 读取
.autopilot/index.md;没有索引时退回读decisions.md和patterns.md - 探索代码库,识别技术栈、现有模式、测试框架、可复用模块
- 在状态文件中写入:
## 目标## 设计文档## 实现计划## 验证方案
- 设计产物至少包含:
- 目标
- 技术方案
- 文件影响范围
- 风险评估
- 测试策略
- 1-3 个真实场景验证步骤
- 必须启动 plan reviewer 子代理,最多 5 轮(0 BLOCKER 即放行)
- 完成后设置
gate: "design-approval"并停下等待用户审批
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago Changed 6f0301f22b84
- 1mo ago First seen · 153 lines · 77 tokens per session scan A d48e05e283b0
autopilot-codex is a skill published in the GitHub repository strzhao/autopilot (32 stars, last pushed 7d ago), licensed MIT. It adds 77 tokens to every session and 1,540 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
fix-bug
Diagnoses and fixes bugs using a test-first workflow (reproduce, diagnose, fix). Use this skill whenever the user reports a bug, describes unexpected behavior, or asks to investigate or fix a defect. Even if the user says "something's broken", "this isn't working", "fix this bug", or "why does X happen", load this…
work
Execute spec.json (plan mode) or review.findings.json (fix-findings mode) by dispatching subagents per chunk. Triggers on "work on", "implement", "execute plan", "carry on", "continue".
plan-creation
Research codebase, validate external claims, and emit a work-ready spec.json for a new planning session. Single-pass creation with integrated validation via Context7 and locator/analyzer agents. Triggers on "create plan", "plan for", "write a plan". For exploratory requests where the user is unsure what to build…
astronomer-airflow
Work with Airflow 3.x on Astronomer. Start/stop, create/trigger DAGs, read logs, run tests. Triggers on "astro", "airflow", "start airflow", "trigger dag", "dag logs".
debug
Debug issues with iterative fix loop. Gathers problem description, investigates, then enters fix-verify cycle. Triggers on "debug", "fix this", "troubleshoot". Use when the goal is to fix a specific reported issue. For exploration or new features, use brainstorm or plan-creation.
tool-connector
Connect any tool you use at work to your agent — including internal company tools, custom-built systems, deployment portals, incident trackers, internal knowledge bases, HR systems, and commercial tools like Slack, Confluence, Jira, Linear, GitHub, Outlook, Datadog, and PagerDuty. Use when the user wants to set up a…