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/i-stack/ai-coding-kit/cross-model-reviewnpx skills add i-stack/ai-coding-kit --skill cross-model-reviewgit clone --depth 1 https://github.com/i-stack/ai-coding-kitWhat 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 | $0.00110 | $0.00986 |
| Opus 5 | $0.00055 | $0.00493 |
| Sonnet 5 | $0.00022 | $0.00197 |
| Haiku 4.5 | $0.00011 | $0.00099 |
Grade A, and why
cross-model-review 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.
What it actually says
Cross Model Review
强制入口
命中本 skill 时，必须先完整阅读 references/cross_model_review.md 并按其中条款执行。
- 不得以 preamble、Cursor 规则摘要或其它二次摘要代替该文件全文。
- 本 skill 是
plan-grill的 Act 2；plan-grill 锁定 PLAN.md 后接力本 skill。
五条核心规则
- [CMR-001] 自动发现 reviewer：直接探测 codex/gemini/claude 三个 CLI 的可用性、版本、non-interactive 与只读模式支持（可用
command -v <cli>+<cli> --version，本仓库可选辅助脚本为skills-engineering/scripts/detect-review-clis.sh）。可用 provider < 2 时停止并提示安装，不伪造 cross-model。 - [CMR-002] 推荐组合 + 用户选择：从可用 CLI 中推荐两个不同 provider 的组合（如 codex + gemini），让用户确认。不静默替用户选死。
- [CMR-003] reviewer 只读：每个 reviewer 必须以只读模式运行——codex 用
-s read-only，gemini 用--approval-mode plan，claude 用--permission-mode plan。reviewer 不写代码，只输出VERDICT: APPROVED或VERDICT: REVISE+ 具体修改建议。 - [CMR-004] 主 agent 仲裁：主 agent（Claude/Codex，视宿主而定）是最终仲裁者。每轮必须收集所有已选 reviewer 的 verdict；reviewer 原始输出、中间输出和交付日志必须保存在当前项目根目录下（推荐
.plan-reviews/<date>-<slug>/raw/），不得用/tmp作为 reviewer 输出缓冲。只有全部APPROVED才能收敛，任一REVISE都必须仲裁并进入修订/下一轮。采纳有证据的批评，拒绝不成立的批评并写明理由，记录进PLAN-REVIEW-LOG.md。 - [CMR-005] MAX_ROUNDS + deadlock：到 MAX_ROUNDS（默认 5）仍不收敛时，输出 deadlock——列出每个未决点 + 主 agent 的反立场，交给用户裁决。禁止假装 approved。
细则见 references/cross_model_review.md。登录限流场景的完整运行样例见 examples/regression-login-rate-limit.md。
何时加载
- 默认触发：用户说
cross-model-review/cross review/ "对抗审查" / "让两个模型审计划" / "model debate" / "stress-test PLAN.md" / "review PLAN.md" / "让 Gemini/Codex/Claude 审一下计划"。 - 接力 plan-grill：plan-grill 锁定 PLAN.md 后，用户说"让另一个模型审查"则加载本 skill。
- 跳过：没有 PLAN.md（先跑 plan-grill）；trivial 改动；用户明确"直接实施"。
与相邻 skill 的分工
| Skill | 分工 |
|---|---|
plan-grill（PG-001~004） |
盘问锁定 PLAN.md（Act 1） |
| cross-model-review（本 skill） | 跨模型对抗审查 PLAN.md（Act 2） |
problem-analysis（PA-001~003） |
问题审查，先于 plan-grill |
epistemic-integrity（GR-011~013） |
主 agent 仲裁时的真值接地纪律 |
What ships with it
8 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.
- AGENT-BRIEF.md 1.9 KB
- examples/regression-login-rate-limit.md 13 KB
- i18n/en-US/references/agent_brief.md 2.3 KB
- i18n/en-US/references/cross_model_review.md 16 KB
- i18n/en-US/references/out_of_scope.md 1.3 KB
- i18n/en-US/references/skill.md 3.8 KB
- OUT-OF-SCOPE.md 1.0 KB
- references/cross_model_review.md 14 KB
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 · 41 lines · 110 tokens per session scan A 8cd783d7ed14
cross-model-review is a skill published in the GitHub repository i-stack/ai-coding-kit (3 stars, last pushed 2d ago), licensed MIT. It adds 110 tokens to every session and 986 once invoked, about $0.0006 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.
Other skills, from other repositories
extension-creator
Create AiderDesk extensions by setting up extension files, defining metadata, implementing Extension interface methods, and updating documentation. Use when building a new extension, creating extension commands, tools, or event handlers.
githits-plugin-maintenance
Internal repository-maintenance skill for GitHits cross-host plugin and Agent Skill surfaces. Use only while working in the githits-cli repository when changing skills, agent guidance, plugin or marketplace manifests, Gemini or Cursor extensions, MCP transport metadata, root release metadata, plugin generation, or…
skillbuddy-skill-creator
通过自然语言对话创建或更新可由多个本机 AI agent 使用的 Skill，并检查重复、触发条件、资源依赖与跨平台兼容性。用户在 SkillBuddy 中要求新建、完善、迁移或适配 Skill 时使用。.
githits-mcp
Use GitHits MCP as an OSS context layer when a task involves open-source packages, frameworks, SDKs, libraries, developer tools, package docs, repository source, examples, planning, research, vulnerabilities, changelogs, dependency graphs, or upgrade-review evidence. Prefer it before relying on model memory or generic…
organize
Open the Cross-Code Organizer (CCO) dashboard — view and manage all memories, skills, MCP servers, hooks, and configs across scopes.
add-agent-property
Add a new property to the AI agents database. Use when the user wants to add, create, or introduce a new column, property, field, or feature to track across all agents in the comparison matrix. Handles all four required steps - database updates, groups.json, table display, and GitHub issue templates.