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 BackToCimaCoppi/Praxis --skill ask-codexgit clone --depth 1 https://github.com/BackToCimaCoppi/PraxisWrote 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/backtocimacoppi/praxis/ask-codex)<a href="https://agentmods.dev/skills/backtocimacoppi/praxis/ask-codex"><img src="https://agentmods.dev/badge/skills/backtocimacoppi/praxis/ask-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/backtocimacoppi/praxis/ask-codex"><img src="https://agentmods.dev/badge/skills/backtocimacoppi/praxis/ask-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 | $0.00070 | $0.04265 |
| Opus 5.5 | $0.00028 | $0.01706 |
| Sonnet 5.5 | $0.00014 | $0.00853 |
| Haiku 4.5 | $0.00007 | $0.00426 |
Grade A, and why
ask-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 29d 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 — 307 lines — stays where its author put it; the contents beside it link to each section on GitHub.
codex 咨询 / 派任务（ask-codex）
通过 codex exec 一次性非交互命令，让 GPT-5.5 承担第二意见、审查、或直接执行任务。
设计哲学：人保留分工权和最终决策权。AI 不做自动路由——用户不说触发词，主线程自己处理。用户说触发词，主线程严格按触发词对应模式执行。
与可持续对话子代理的根本差异：当前运行时的可持续对话子代理，子代理是可持续对话的进程，派模式能"先探索汇报计划、等确认再动手"。codex 不是子代理，是一条一次性非交互命令行——跑完即结束，没有"跑到一半停下来等确认"这回事。因此本 skill 的派模式是一次性直接执行（任务+护栏一次性打包，执行完由主线程做事后 git diff 验收），不是"探索计划→确认→继续"的两阶段模式。
0. 与 codex-review / adversarial-review 的边界（进入本 skill 前必查）
codex 在这套技能体系里有三个入口，互不替代：
| Skill | 时机 | 前提 | 输出形态 |
|---|---|---|---|
adversarial-review |
开工前，方案/高风险代码待定 | 无固定前提 | 多模型评审 + 主线程裁决报告 |
codex-review |
施工完成后 | 必须有已冻结蓝图 + 决策表（或蓝图退化基线） | 强制 A/B/C 三级分类 + 证据闸 + 分流台账 |
| 本 skill（ask-codex） | 任意时刻 | 无前提，随问随答 | 自由文本结论 / 直接执行 |
强制检查（触发本 skill 前）：问题是不是"核验代码是否照已冻结的蓝图/决策表施工"？
- 命中 → 停止，改用
codex-review。它的证据闸和三级分流协议专门防止"把已拍板决策误判成 bug、又被照改"，本 skill 没有这套保护，绕开它会架空那套纪律。 - 没命中（轻量咨询、非代码问题、没有冻结真值可核对、或直接派任务执行）→ 才用本 skill。
方案/高风险代码的开工前评审同理指向 adversarial-review，不进本 skill。
1. 模型映射
| 模式 | 触发词 | codex 参数 |
|---|---|---|
| 问 codex（默认） | 问一下codex、问一下gpt、问一下gpt5.5 | -m gpt-5.5，reasoning 默认 xhigh（config 默认即是，无需显式传） |
| 派 codex | 让codex去做XX、让codex直接XX、让gpt去做XX、让gpt直接XX | 同上 |
- 用户未指定降级时，默认 xhigh（超强思考）
- 用户说"快速 / 省着点跑" → 降级
-c 'model_reasoning_effort="medium"' - 用户不说触发词 → 不调 codex，主线程自己处理
- 触发词模糊时 → 直接问用户："你想问模式还是派模式？"
成本提示：xhigh 质量最高也最慢（常 60–180s，复杂任务更久），调用必须走后台（§5）。
2. 模式路由
用户消息
│
├─ 不含触发词 → 主线程直接处理
│
└─ 含触发词
│
├─ 先过 §0 边界检查 → 命中 codex-review / adversarial-review 范围 → 转对应 skill，不进本 skill
│
└─ 未命中
│
├─ "问一下" / "怎么看" / "审查" / "评审"
│ → 问模式（§3）
│
└─ "让XX去做" / "让XX直接" / "派XX"
→ 派模式（§4）
3. 问模式
用途：让 codex 回答一个问题（设计审查、技术分析、方案评估、代码片段审查等）。codex 只读不改，输出结论到文件，主线程消化后回答用户。
沙箱：-s read-only（物理上写不了任何东西）。
3.1 打包指南
方法论复用 §8 引用的证据包制作规范（RAM 框架：R 谁读 / A 干什么 / M 最小充分集）。槽位：
[具体问题] — 一句话，可被独立回答，不依赖隐含上下文
[文件引用] — 需要读的代码/文档文件，精确到方法/行范围
格式：文件路径（只读：方法名/行范围）
必须附带禁入声明："只读以上列出的文件，不要自行查找其他文件"
[背景事实] — 回答此问题必须知道的关键事实（先浓缩再写入，不贴超过 5 行原始代码）
[约束条件] — 从 AGENTS.md 摘出适用条款，内嵌原文，不写"参考 AGENTS.md"
[判断标准] — 什么算好答案
[外部信息] — 非代码的参考信息（如有）
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.
- 29d ago Changed · -1 lines 98d3f2d4cf95
- 1mo ago First seen · 308 lines · 70 tokens per session scan A 8c39ddec1aac
ask-codex is a skill published in the GitHub repository BackToCimaCoppi/Praxis (6 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 70 tokens to every session and 4,265 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 2026-08-31.
Other skills, from other repositories
prompt-master
Generates optimized prompts for AI tools. Activates only when the user explicitly asks to write, fix, improve, or adapt a prompt for a specific AI tool (LLM, Cursor, Midjourney, image AI, video AI, coding agents, etc.). Does not activate for general conversation, coding tasks, document writing, or other…
browser-tools
Safety rules over the upstream agent-browser skill (scraping ethics, auth, request pacing, local URLs). URL policy is the Claude Code sandbox's job when the operator enables it. Use for browser automation.
rag-retrieval
Retrieval-Augmented Generation patterns for grounded LLM responses. Use when building RAG pipelines, embedding documents, implementing hybrid search, contextual retrieval, HyDE, agentic RAG, multimodal RAG, query decomposition, reranking, or pgvector search.
architecture-patterns
Architecture validation and patterns for clean architecture, backend structure enforcement, project structure validation, test standards, and context-aware sizing. Use when designing system boundaries, enforcing layered architecture, validating project structure, defining test standards, or choosing the right…
auto
Intent-classified router, the front door to OrchestKit and the DEFAULT entry point for any goal-shaped request. Classifies a plain-English goal and routes it to the right specialist skill. Routing is never overhead, so use it even when the target skill seems obvious; skip only when already executing inside another…
langgraph
LangGraph 1.x (LTS) Python workflow patterns for state management, delta channels, resilience (node timeouts, error handlers, graceful drain), routing, parallel execution, supervisor-worker, tool calling, checkpointing, human-in-loop, streaming (v2 format), subgraphs, and functional API. Use when building LangGraph…